How Emotions Influence Consumers' Perception of Credibility and Trust in CSR Communication

Home , Emotionality

Department of Informatics and Media

Digital Media and Society

Two-Year Master’s thesis

The Age of Emotionality? – How emotions influence consumers’ perception of credibility and trust in CSR communication

Student: Anika Reupsch

Supervisor: Göran Svensson

Spring 2017

Abstract

Companies around the world are using different strategies for their corporate social responsibility (CSR) communication, but finding an appropriate strategy to enhance trust and credibility on the consumer side remains challenging. The constitutive aspect of emotions in CSR communication has long been overlooked. Therefore, this study investigates the influence emotions in CSR communication have on the credibility and trust consumers have in a firm’s CSR.

Quantitative research with group division was conducted. The online survey used authentic CSR communication content of a German trading company, which was classified in two categories: low emotional appeal and high emotional appeal. Statistical tests for group differences, correlation, factor analysis and multiple regression were carried out to study the influence of emotions on credibility and trust.

The applied test statistics for comparing groups differences have shown a significant difference for guilt and regret, which are reduced for the group exposed to high emotional CSR content. Similarly, comprehension is enhanced, whereas sincerity is reduced for high emotional content. Above that, it was proved that positive emotions have a positive effect on perceived credibility for both text- image and moving image CSR content. However, negative emotions influence perceived credibility only for the text-image content. To the contrary, the analysis for moving image content revealed another dimension formed by interest and compassion, that positively influences perceived credibility. Regarding the dimensions of credibility, it was shown that understandability is negatively influenced by negative emotions, whereas perceived truth, sincerity and normative appropriateness were enhanced by positive emotions.

The findings support, that emotions should be taken into account when analyzing the effects of CSR communication. Companies, but also academic research needs to address that credibility of a company and its CSR communication is influenced by the consumers’ emotions, positively as well as negatively. Hence, the creation of CSR communication content needs to carefully address which medium to choose, which emotions are desirable and which should be avoided, when communicating CSR to the consumer.

Keywords: corporate social responsibility, corporate social responsibility communication, corporate communication, public trust, credibility, emotions, media selection

Table of contents

List of figures ...... 6

List of tables ...... 6

1. Introduction ...... 8

1.1. Research purpose und questions ...... 8

1.2. Thesis disposition ...... 10

2. Literature review...... 11

2.1. Setting the scene – approaches to CSR ...... 11

2.1.1. The evolvement of CSR ...... 11

2.1.2. Concepts and Definitions ...... 12

2.1.3. The stakeholder approach to CSR ...... 14

2.2. CSR communication ...... 15

2.2.1. Foundations of CSR communication ...... 15

2.2.2. Challenges in CSR communication ...... 16

3. Theoretical Framework...... 18

3.1. Trust and Trustworthiness ...... 18

3.2. Credibility and Legitimacy ...... 20

3.3. Emotions and Emotional appeal ...... 22

3.4. Theoretical discussion ...... 27

4. Methodology ...... 30

4.1. Research design ...... 30

4.2. Case Study approach ...... 31

4.3. Data collection ...... 33

4.4. Measuring emotions ...... 34

4.5. Measuring credibility and trust ...... 36

4.6. Statistical considerations ...... 38

4.6.1. Population and sampling ...... 38

4.6.2. Reliability and validity ...... 38

4.7. Statistical tests ...... 39

4.7.1. Hypothesis testing and significance level ...... 40

4.7.2. Likert type items and Likert scale ...... 41

4.7.3. Comparing groups with statistical tests ...... 41

4.7.4. Bivariate Correlation ...... 42

4.7.5. Factor Analysis ...... 43

4.7.6. Multiple Regression ...... 44

4.8. Ethics ...... 46

4.9. Reflection ...... 46

5. Results ...... 48

5.1. Testing group differences ...... 48

5.1.1. Differences in mediated emotions ...... 48

5.1.2. Differences in perceived credibility ...... 50

5.2. Correlation between emotions and credibility ...... 51

5.3. Factor analysis ...... 55

5.3.1. Factor analysis for data reduction ...... 55

5.3.2. Correlation between extracted factors and dimensions of credibility ...... 60

5.4. Multiple regression ...... 63

5.4.1. Regression model for perceived credibility and trust after reading the text ...... 63

5.4.2. Regression model for perceived credibility after watching the video ...... 64

6. Analysis ...... 66

7. Conclusion and Discussion ...... 72

7.1. Future research ...... 73

7.2. Practical implications ...... 74

8. References ...... 77

Appendix ...... 82

List of figures

Figure 1: The pyramid of CSR by Carroll (1991) ...... 12 Figure 2: The mediating role of moral emotions in corporate environmental irresponsible actions ...... 26 Figure 3: The mediating role of moral emotions in corporate environmental responsible actions 27 Figure 4: Hypothesized model for the quantitative analysis ...... 28 Figure 5: The Geneva Emotion Wheel (GEW) by Scherer (2005) ...... 35 Figure 6: Boxplots for Guilt and Regret ...... 50 Figure 7: Scree plots for factor analysis ...... 59

List of tables

Table 1: Elements and types of public trust ...... 19 Table 2: Factors of public trust ...... 19 Table 3: Components and functions of emotions ...... 23 Table 4: Other types of affect and explanation...... 24 Table 5: Criteria for assessment of emotional appeal (own compilation) ...... 32 Table 6: Classified CSR communication content (own compilation) ...... 33 Table 7: Factors and measurement variables for credibility and trust ...... 37 Table 8: Cronbach's Alpha for text-related credibility ...... 39 Table 9: Cronbach's Alpha for video-related credibility ...... 39 Table 10: Mann-Whitney U-Test Output for Guilt (text-related)...... 49 Table 11: Mann-Whitney U-Test Output for Regret (video-related) ...... 49 Table 12: Significant difference between G1 and G2 for credibility and trust (text-related) ...... 51 Table 13: Significant difference between G1 and G2 for credibility and trust (video-related)..... 51 Table 14: Strength and direction of the correlation between credibility and individual emotions (text-related) ...... 52 Table 15: Strength and direction of the correlation between credibility and individual emotions (video-related) ...... 52 Table 16: Spearman correlation for Credibility and Emotions (text-related) ...... 53 Table 17: Spearman correlation for Credibility and Emotions (video-related) ...... 54

Table 18: Correlation for dimensions of credibility and emotions (text-related) ...... 54 Table 19: Correlation for dimensions of credibility and emotions (video-related) ...... 55 Table 20: KMO and Bartlett's Test for text-related emotions ...... 56 Table 21: KMO and Bartlett's Test for video-related emotions ...... 56 Table 22: Factor reduction for text-related emotions ...... 57 Table 23: Factor reduction for video-related emotions ...... 58 Table 24: Extracted factors and representative variables (text-related) ...... 60 Table 25: Extracted factors and representative variables (video-related) ...... 60 Table 26: Correlation between extracted factors and dimensions of credibility (text-related) ..... 61 Table 27: Correlation between extracted factors and dimensions of credibility (video-related) .. 62 Table 28: Computed coefficients for the regression model (text-related) ...... 64 Table 29: Computed coefficients for the regression model (video-related) ...... 65

1. Introduction

It seems like our world has changed – emotion-driven people can become president of very powerful democratic states1, people publicly talk about “alternative facts”2 and media postulate that we might have entered a new “post-factual” era (Preston 2016).

Emotions are, of course, nothing new. Since the very beginning emotions drive human behavior. However, for many years rational thinking was (and is) considered as one of the most important thinks, that distinguish us – the human beings – from being an animal.

And this is, where my journey starts: I want to know: How much do emotions influence our evaluation and perception? Since this in general would be a huge area, I decided to take it to the communication realm, namely to organizational communication. By linking it to another pressing issue – sustainability, I will investigate corporate social responsibility (CSR) communication, its emotional appeal and the emotions it creates for the consumer. Which reactions and perceptions do emotions create for the consumer and how does this influence perceived credibility and trust?

This thesis will study different forms of authentic CSR communication content using a case study approach. It investigates the CSR communication of a German trading company, which uses different channels and platforms to communicate CSR to its stakeholders. For this two types of media have been selected: text-image content, that combines text and images and is available on different online platforms of the firm; and moving image content, that provides video content that deals with CSR issues. Both types of media are publicly available online and provide opportunities to inform oneself about the CSR engagement and activities of the firm. What was excluded from the analysis are print products and TV commercials.

1.1. Research purpose und questions

The concept of CSR occupies a substantial amount of attention in theory, as well as in practice. Although the concept of responsible business engagement is not a new one, CSR became a buzzword, especially in newer management literature.

1 This does refer to the new president of the United States of America, Donald Trump, who is known for acting emotional rather than rational (Hermida 2016) 2 The Counselor to the President Kellyanne Conway spoke about alternative facts given by the White House Press Secretary Sean Spicer on the number of attendees to the inauguration of US president Trump.

Though being an interdisciplinary research field, CSR research is still dominated by management and business studies. What is under-researched and maybe undervalued is its influence and impact on communication processes by being part of the organizational communication.

Subjects of communication studies are constituted by the reconstruction, description and analysis of public communication processes (Raupp, Jarolimek, and Schultz 2011). From a scientific view organizations are relevant communication actors to study, since they directly communicate with semi-publics (stakeholders), and at the same time provide topics and information for the media. However, in recent research a growing focus on the communicational and medial processes in CSR could be identified (Raupp, Jarolimek, and Schultz 2011). Especially the field of Public Relations (PR) research engages more and more with CSR, understanding it as a subfield of PR (Röttger, Preusse, and Schmitt 2014; Lock 2016).

With regards to emotions, CSR communication initiatives are considered the most relevant, since the interplay of plain facts and emotions - e.g. how much a company spends for social projects or how a video clip illustrates the social engagement of a firm by showing happy and well-educated kids – is very common in CSR communication and many firms make use of it.

Though several articles investigate the role of credibility and trust consumers put into a firm’s CSR activities (Bentele and Seidenglanz 2008; Bentele and Nothhaft 2011), only a few investigate the role of emotions in CSR communication and its impact on issues like trust and credibility (Pérez and Rodríguez del Bosque 2015; Farache and Perks 2010). The purpose of this paper is it, thus, to study how emotions and emotional appeal in CSR communication influence the customers in their perception of the company’s credibility and the amount of trust they put into the firms CSR. Is the influence of emotions indeed stronger than argument?

By applying empirical methods this thesis seeks to answer the following research questions:

RQ 1: Do the emotional appeal of CSR content and the consumers’ emotions influence the credibility and trust consumers put into a firm’s CSR communication?

RQ 1.1 How can emotions and credibility of CSR communication be measured? RQ 1.2 Do different forms of CSR content mediate different emotions? RQ 1.3 To which extend do the consumers’ emotions influence the perceived credibility of a company?

1.2. Thesis disposition

The thesis starts by exploring the research field to which it will contribute: CSR, particularly CSR communication. This is done with an extended literature review including the evolvement of CSR, different concepts and definitions that have been developed. The stakeholder approach, which most CSR literature makes use of nowadays is also explained shortly.

In chapter 3, a theoretical framework for the empirical analysis is developed. The theory part appears two-folded using theories on trust and credibility on the one hand and theories on emotions and emotionality on the other hand. After approaching different theories, the conceptual model for the thesis is developed.

After setting the theoretical framework, the method part is addressed. The research strategy is outlined and described. The case study approach is explained and the used CSR content gets classified regarding its emotional appeal, before the data collection process is illustrated. The study uses quantitative methods to tackle the research question. To successfully measure emotions and credibility of CSR communication, a measurement model for CSR communication content is developed (see RQ 1.1). For the further analysis of the collected data, some statistical considerations are made and the applied statistical tests are briefly explained. Ethics and a short reflection on the chosen methods are also presented.

The thesis continues, with the results and analysis part. The findings from the different statistical tests are presented and analyzed for their impact on the conceptualized model.

This is done by statistically assessing the two sub-research questions:

RQ 1.2 Do different forms of CSR content mediate different emotions? RQ 1.3 To which extend do the consumers’ emotions influence the perceived credibility of a company?

The thesis ends with a summary of the research outcomes and a discussion on future research and practical implications.

2. Literature review

Though CSR only recently became a buzzword in modern economy, the media and politics; several researchers have shown that the majority of US companies has had CSR activities and/or officers in the 1970s already (Eilbirt and Parket 1973; Våland and Heide 2005; Pedersen 2015). Above that the origins of CSR in the form of business ethics can be traced back to Greek philosophers, the Bible and the Middle Ages (Pedersen 2015). What becomes clear here, is that CSR not only is a vague concept with several definitions and overlapping concepts, but in addition it is by far not a new concept – but has a long history.

To avoid confusion, I will draw out the history and evolvement of CSR shortly, before I come to CSR and CSR communication as concrete concepts, presenting definitions of how CSR and CSR communication will be understood in this thesis. At the end of this chapter I will present the challenges CSR communication faces.

2.1. Setting the scene – approaches to CSR

2.1.1. The evolvement of CSR

The long history of CSR began with Aristotle, who in his work “Politics” philosophized about administration and control, which should respect and create benefit for the people, rather than just accumulate property.

With increasing size and power of private corporations CSR became an issue during the Industrial Revolution, while still being mixed with the individual social responsibility of the firm owner (Cannon 1994). In the 19th century the structure of companies changed, professional managers emerged and ownership and control became separated. Individual responsibility, thus, became more and more a corporate social responsibility. The growing importance of companies not only in economy, but also as a political and social actor in the following years, together with pressing issues like climate change and working conditions, slowly but steadily pushed CSR to the public agenda. (Pedersen 2015)

Despite its historical roots, CSR is often seen as a post-World War II phenomenon, with the book “Social Responsibilities of the Businessman” by Howard R. Bowen’s (1953) being understood as the beginning of today’s CSR discussions (Pedersen 2015).

CSR, as any other concept, is shaped by its societal surrounding. This includes trends and changes in politics, economy and society. As it will be further explained at the end of this chapter, two trends became central for the growing relevance of CSR: globalization and the growing societal expectations from stakeholders.

2.1.2. Concepts and Definitions

As it was outlined above, the concept of corporate social responsibility (CSR) has a long tradition, especially in the U.S. Consequently, there exist several conceptions from different disciplines, while still being dominated by business and management studies.

As a basic concept the pyramid by Carroll (1991) is widely used. The illustration draws out four different areas of responsibility: economic responsibilities, legal responsibilities, ethical responsibilities and philanthropic responsibilities.

Figure 1: The pyramid of CSR by Carroll (1991)

The first two levels would be obligatory for a company to survive. The third level, ethical responsibility, would not be obligatory, but is expected by society. The fourth level of philanthropic responsibility would be totally voluntary, though desired by society. What the pyramid lacks, however, is the differentiation into social and ecological problems. Current issues such as global

working conditions, climate change and sustainable business practices are not present and hard to classify into these four categories.

Other authors see the concept more restricted to the societal responsibility of a company, the aspect of voluntariness and add a sustainability component to it (Raupp, Jarolimek, and Schultz 2011). Following their idea CSR actions must have a connection to the core business of the company, be implemented totally voluntary and, finally, be directed (at least implicitly) towards the sustainable use of resources.

However, the International Organization for Standardization with its ISO 26000 standard (2010) defines social responsibility as “responsibility of an organization for the impacts of its decisions and activities on society and the environment, through transparent and ethical behaviour that:

• contributes to sustainable development, including health and the welfare of society; • takes into account the expectations of stakeholders; • is in compliance with applicable law and consistent with international norms of behaviour; and • is integrated throughout the organization and practised in its relationships” (p.3)

What is highlighted in this definition are ethical and transparent behavior, the environment and the society, but also the expectations of the stakeholders are explicitly mentioned here. Thus, the important role of the stakeholders in a firm’s social responsibility is acknowledged and even became part of the definition.

The European Commission was one of the pioneers in the promotion of CSR through public policy initiatives. In 2001, they released a Green Paper and subsequently established the European Multistakeholder Forum on CSR. With the Europe 2020 strategy, however, the Commission made a commitment to further promote and work on CSR. When in 2011 they released their new policy “A renewed EU strategy 2011-14 for Corporate Social Responsibility” they slightly changed their core definition of CSR from “a concept whereby companies integrate social and environmental concerns in their business operations and in their interaction with their stakeholders on a voluntary basis” (European Commission 2001) to a more general approach defining CSR as “the responsibility of enterprises for their impacts on society” (European Commission 2011). Surprisingly, the voluntariness was erased from the new definition. What it is added to the “new”

definition, however, is the respect for applicable legislation and collective agreements between social partners. To fully meet their corporate social responsibility the Commission asks for processes to “integrate social, environmental, ethical, human rights and consumer concerns into their business operations and core strategy in close collaboration with their stakeholders, with the aim of:

• maximising the creation of shared value for their owners/shareholders and for their other stakeholders and society at large; • identifying, preventing and mitigating their possible adverse impacts.” (European Commission 2011:6)

From all these different definitions, it gets clear that CSR does not have one definition. With all these diverse and changing concepts, CSR has often been accused of vagueness, ambiguity and a lack of clarity (Pedersen 2015:12). When considering the multitude of overlapping concepts such as corporate citizenship, business ethics or corporate accountability the confusing picture is complete.

However, though the struggle for a proper definition goes on, most (contemporary) definitions can be melted down to the following characteristics (based on Pedersen 2015):

1) CSR is multidimensional and does include social and environmental issues 2) CSR is voluntary and goes beyond legal obligation (though this concept is challenged as the “new” definition by the European Commission shows) 3) CSR is about stakeholders.

For this paper, it will be worked with this three-folded compressed definition. The most important issue to address will be the stakeholder, the interrelationship between business and society. To understand this interrelationship, scholars have to analyze the groups and individuals who form the stakeholders of the company in the society.

2.1.3. The stakeholder approach to CSR

The theoretical foundation of CSR is as diverse as the concept itself, being approached from different academic disciplines. However, nowadays the most dominant approach to CSR is the stakeholder approach (Pedersen 2015; Freeman 2010; Carroll 1991).

Stakeholder are understood as individuals or groups, that have a “stake” in the company. More precisely, stakeholders have a certain interest, that is that the company creates a certain value for them (Parmar et al. 2010). This relationship is two-sided. As such the company affects and is affected by its stakeholders (Pedersen 2015: 13). Following these ideas, the company has to create value for a wide range of stakeholders: employees, suppliers, customers, media, local communities, NGOs, trade unions, investors etc. Much of the CSR literature argues that a company that tries to fulfill the stakeholders’ economic, social and environmental demands, will lead to the highest long- term value for both businesses and society (Pedersen 2015: 14).

This thesis, however, will concentrate on one stakeholder group in greater detail: the consumer. Consumers naturally have a great market power, since they decided to buy products from a certain firm and consequently provide the firm’s revenue. The survival and success of a company is as such heavily dependent on the consumer (Carrington and Neville 2015).

Simultaneously with the new “ethical consumerism”, which describes the growing consumer awareness for ethical and environmental issues in production and (over-)consumption of goods, companies changed their marketing and communication strategies by putting more focus to CSR (Carrington and Neville 2015: 187-188, 197-199). Interestingly, only a few ethically minded consumers follow their consumption intentions. In 2005, 89% of the UK consumers stated to have ethical concerns (FUTERRA Sustainability Communications Ltd 2005), but only 30% of UK consumers intended to purchase ethical products and only 3% actually did so, following figures from 2001 (Lazzarini and Carvalho de Mello 2001). Though, this is some years ago and awareness for conditions of production and the environment have grown, it can be assumed that this gap still exists. Several explanations were given for this “ethical purchase gap”. One of them views the role of marketing and communication as critical for the ethical consumption paradox (Lazzarini and Carvalho de Mello 2001: 197-199). In this sense, CSR communication has a great impact on consumer behavior.

2.2. CSR communication

2.2.1. Foundations of CSR communication

CSR communication is studying corporate responsibility from media and communication studies perspective. Rather, than studying how social reputation is an essential source of capital for the

business which can generate heavy return on investment and thus potentially generates a competitive advantage, its studies corporate responsibility where it becomes a real tangible phenomenon: in popular culture and in the (mass) media (Bentele and Nothhaft 2011).

Studying CSR from a communication studies perspective is relevant due to several aspects (see Raupp, Jarolimek, and Schultz 2011):

1) Communication studies asks for the communicative and media conditions, in which a company acts and how their behavior is influenced by political, economic, social and cultural condition. The conception of legitimacy, credibility, reputation and moral are central for CSR communication research. 2) Furthermore, CSR communication research investigates how CSR communication practices are dependent on the firms’ organizational features, such as size, internationalization, conception of social responsibility and industry. 3) CSR communication has different influence on the public perception of corporate responsibility and legitimacy. Who are the actors in the discourse and how can the public discourse be influenced by companies and the society? 4) CSR communication should not be confused with advertising. On its operative level, it is closely related to PR, while it is more complex from the strategic level and thus needs further research from organizational communication. 5) The “communicative turn-around” in CSR research calls for an interdisciplinary interconnection and a higher involvement of communication studies.

In line with the above-mentioned arguments for the investigation of CSR from a communication studies perspective, this paper strives to contribute to this field by studying authentic CSR communication and its influence on the public perception of trust and credibility.

2.2.2. Challenges in CSR communication

Two phenomena have been identified as the current main drivers of CSR by Pedersen (2015): globalization and growing societal expectation/pressure. Both are understood as challenges to CSR communication.

In consequence to what we call “a globalized world” or “global village” (Giddens 2007) both society and economy are more connected than ever. This growing interconnectedness between

people, economies and cultures means that local events can have global consequences and vice versa (Pedersen 2015). Firms cannot hide away from the responsibility they might have for happenings abroad, though it might be hard to keep an overview in globalized and complex supply- chains. In the same manner, communication was globalized. The advent of the internet and the emergence of social media, put more power to the stakeholder, who is no longer dependent to the information he gets by the company or traditional media.

Many researchers, however, support the idea that it is mainly the increasing pressure of the stakeholders, that makes CSR a pressing issue for companies to address and fulfill (see 2.1.3 on the stakeholder approach). It is argued that companies need to respond to new social and environmental demands, which are communicated through “hyperactive media” (Pedersen 2015). Otherwise, they will suffer from negative media and grass root activism or consumer sanctions and negative long-term impacts on image, reputation and profit.

With the new media and societal consciousness, it is no longer enough to have a CSR strategy, it is essential to communicate this CSR engagement to the public. More precisely the main challenge of CSR today is how to effectively communicate with the consumer in a constantly changing and emotionalized environment. The challenge for companies is it to stay ahead of criticism in the public discourse. The communicative process is about retaining legitimacy, gaining trust and being valued as a credible and reliable source of information.

Consequently, in the next chapter a theoretical framework for the analysis of CSR communication will be developed by addressing the concept of trust, trustworthiness, credibility, legitimacy and the role of emotions and emotional appeal.

3. Theoretical Framework

To address the issue of emotions in CSR communication this chapter builds up a framework using concepts such as credibility and trust from business and communication studies and some considerations on emotions grounded in psychology, social and communication studies.

From a communication and media studies perspective CSR is a double phenomenon: It is constituted by the company’s view of social responsibility and simultaneously by the society’s perception of corporate responsibility. Bentele and Nothhaft (2011) identified two central mechanisms as crucial for the analysis of CSR communication: trust and credibility. Since these two concepts are often intertwined in the literature (Rouner 2008; Lock 2016), it is important to analyse the two concepts before using them for the assessment of the collected data.

3.1. Trust and Trustworthiness

Referring to Luhmann (2000) trust can basically be defined as a (communicative) complexity- reducing mechanism, a risky prior concession, where future events are predicted based on past experience (Bentele and Seidenglanz 2008). Defining trust as a communicative process, it faces the same challenges and changes as communication. Living in a modern globalized society trust (just as communication) is no longer bound to face-to-face communication but rather mediated through a variety of media. As a result, interpersonal (trust to persons) and social trust (trust in a group of people), are more and more accompanied by the so called public trust.

Bentele and Nothhaft (2011) state that mediated communication, such as mass media, journalism, but also PR and corporate communication play a key role in the formation of trust nowadays. The theory of public trust (Bentele 1994; Bentele and Seidenglanz 2008) outlines crucial elements in the process of public trust and describes four distinct types of trust:

Table 1: Elements and types of public trust

Elements of public trust Types of public trust subjects of trust: groups or person, which trust interpersonal basic trust objects of trust: public persons, organisations or public trust in persons social systems that are trusted mediator of trust: PR/Corporate Communication, public trust in organizations Journalism/Media facts and circumstances public trust in social systems Source: Own compilation based on Bentele and Nothhaft (2011)

Additionally, the theory of public trust postulates the existence of factors of trust, described as certain attributes that the subject of trust wants the object of trust to fulfil. Factors of trust get visible in the communicative and general behaviour and actions of persons and organizations alike. If several of these factors are fulfilled or fulfilled distinctively, there is a high chance of building up trust, while on the other hand the absence of these factors results in a destruction of trust (Bentele and Nothhaft 2011: 55). In total seven factors of trust could be identified, which can be grouped in three dimensions.

Table 2: Factors of public trust

Expert dimension Social-normative dimension Social psychological dimension

Professional competence Ethical/ normative behaviour Social behaviour

Problem-solving Sense of responsibility Character competence

Communicative behaviour

Another concept of trust by Mayer, Davis, and Schoorman (1995) highlights the two- dimensionality of trust. The main influencing factor for the trustor is propensity, while the trustworthiness is fundamental for the trustee to build up trust. Since the company has no influence on the trustor’s propensity, it can only influence its trustworthiness. Trustworthiness is constituted by the three factors: ability, benevolence and integrity.

When consulting the Oxford dictionary, trustworthiness is defined as being “able to be relied on as honest or truthful”. It is as such close to the definition of credibility as the “quality of being trusted and believed in” (Oxford Dictionary 2017).

3.2. Credibility and Legitimacy

Being crucial for an effective communication whether it being between two persons, in the mass media or in organizational communication, credibility is understood as one of the most important features recipients can attribute to a communicative process (Lock 2016; Jackob 2008)

In business studies the concept of credibility is often researched in relation to brand reputation and image. Newell and Goldsmith (2001), however, argue that empirical studies on credibility lack a generalizable measurement scale of credibility and are as such, not comparable and not well generalizable. In their work, they compare different factors which have been used previously and generated an empirically verified reliable eight-item Likert-type measurement scale which is based on the two factors: trustworthiness and expertise. Though this scale might be useful for self- assessment of companies, this scale only measures the source credibility of a company, but not the credibility of content or recipient.

Quite similarly several concepts of credibility exist in communication studies (Donsbach 2008). Rouner (2008) for instance, criticizes the lack of separation between the three dimensions of communication (sender, recipient, message) in former credibility research and therefore developed a two-dimensional model which investigated credibility of the sender and credibility of the message alike.

Another concept differentiates between three models of credibility: the “source model”, the “recipient model” and the “experience-model” (Jackob 2008). The source model assigns credibility to the sender. If the sender of the message is credible, the message is more likely to be accepted. This model was also used for the above-mentioned work by Newell and Goldsmith (2001). The recipient model, on the other hand, allocates credibility to the side of the recipient. If the recipient has a positive attitude towards the company’s CSR engagement, he/she will more likely believe in the credibility of the company and its messages. The experience model understands credibility as

a process to experience. As such it was mostly applied in experimental research, for example to study the perception of messages in advertising (Lock 2016: 416).

Following the thoughts of Bentele and Seidenglanz (2011), credibility has to be understood as a sub-concept of trust. Unlike trust, which is given to persons, objects and processes alike, credibility is limited to communicative processes with the combination of person and text (or text only) being the reference. However, credibility is not a characteristic of a text, but an element of a “multipositioned relationship” (p. 2), where the recipient believes that someone or something (e.g. a person, an institution or their communicative products) is credible in relation towards something (e.g. an incident or issue) on the basis of a certain communicative process (e.g. a text, an image or a video clip). In addition to that, it is the timely perspective that distinguishes trust and credibility. Trust is always directed to an unpredictable future and as such a prior concession (see 3.1.), while credibility is bound to the presence (or the past) and can be evaluated immediately.

With regard to CSR communication, Lock (2016) identified credibility as an even more pressing issue for companies to address (p. 414). However, she criticizes this not being an easy task with “credibility” being a rather vague concept in CSR, which is often attributed to the sender only (p.415). She is, thus, requesting a concept which connects sender, receiver and message alike to be able to conceptualize credibility in the complex CSR communication environment. In contrast to Bentele’s concepts (2008, 2011) she views credibility as an independent concept. By developing a legitimacy-based credibility concept based on “the theory of communicative action” by Habermas (1981), she wants to overcome the “legitimacy gap”3. Instead of using strategic action to enhance the company’s reputation Lock, in line with the Habermasian idea of communicative action, is asking for a deliberated discourse, mutual exchange and consensus. Consequently, she argues that the best way for a company to achieve or maintain their credibility is a CSR communication which has a moral legitimacy as its main objective. CSR communication should be based on the four validity claims (Lock 2016: 422):

3 The „legitimacy gap“ is understood as the paradoxon that more communication on CSR did not lead to more credibility in CSR communication, but rather the other way around.

• Understandable: the communication is understandable • Truth: the content communicated is objectively true • Sincerity: the participants and the content are sincere • (normative) appropriateness: the communication is appropriate in its normative-moral context

Following this concept, the CSR communication of a firm is perceived as credible when all participants communicate understandable, truly, sincere and normatively appropriate.

3.3. Emotions and Emotional appeal

Emotions play a more important role than ever in current research. After being excluded from the academic world and deemed as irrelevant for decision-making, in recent years academia undergoes a mayor change, which is often referred to as “the emotional turn”. As Baker and Rowe (2013) highlight, the “emotional turn” marks a paradigm shift in both human and natural science, where emotions are no longer understood as contradicting to rationality, but are rather admitted as a significant constituent to procedural rationality. Social scientists and neuroscientist are, thus, admitting the indispensable role of anticipatory emotion when choosing and acting (Damasio 1994), especially in the “social world” (Barbalet 1998) where it is argued that emotions not only support rational thought and action (e.g. self-interest, decision-making, goal-formation), but that both emotions and rationality are continuous modes of human social behavior (Baker and Rowe 2013).

But what are emotions? – The definition and conceptualization of emotion is a notorious problem in social and behavioral science conglomerated with everyday understanding and misleading scientific definitions that mix emotions with related concepts such as feeling or mood. Using the component process model emotion is defined as “an episode of interrelated, synchronized changes in the states of all or most of the five organismic subsystems in response to the evaluation of an external or internal stimulus event as relevant to major concerns of the organism” (Scherer 2005). Though sounding very complex at first, this definition is easy to break down with the help of the five components that are addressed in the definition.

Table 3: Components and functions of emotions

Emotion component Emotion function

Cognitive component (appraisal) Evaluation of objects and events

Neurophysiological component (bodily System regulation symptoms)

Motivational component (action tendencies) Preparation and direction of action

Motor expression component (facial and Communication of reaction and behavioral vocal expression) intention Subjective feeling component (emotional Monitoring of internal state and organism- experience) environment interaction Source: Own compilation based on Scherer (2005)

With the component model, it is also possible to draw distinctions from related concepts. The term “feeling” for example addresses the subjective feeling component of emotion only, it is thus one component of emotion, related to a person’s internal state and interaction with the environment (Scherer 2005).

Emotion is one type of affection. In the following, distinction from other affective types such as preferences, attitudes, moods, affect disposition and interpersonal stances will be made. Scherer separates these types by assessing several features: event focus, intrinsic appraisal, transactional appraisal, synchronization, rapidity of change, behavioral impact, intensity and duration (Scherer 2005: 700-706, see Appendix 1 for a more detailed overview).

Table 4: Other types of affect and explanation

Type of affect Explanation Preference relatively stable judgement about liking or disliking a stimulus/ object, or preferring it or not over another stimuli/object

stable affect which is based on belief and Attitude predisposition towards specific objects or persons

Mood diffuse affect states with a predominance of a certain subjective feeling that affects experience and behavior of a person, often without being linked to a certain event

personality traits and behavior tendencies Affective disposition with a strong affective core, which describe the tendency of a person to experience certain moods more frequently or to be prone to react with certain types of emotions, even upon slight provocation

characteristic of an affective style that Interpersonal stances spontaneously develops or is strategically employed in the interaction with a person or a group of persons, coloring the interpersonal exchange in that situation Source: Own compilation based on Scherer (2005)

After defining emotions and distinguishing it from other affective states, the terms emotionality and emotional appeal are to be addressed. Emotionality in this work is used as an expression for the observable reaction to emotion. The stronger the emotional response to a stimulus, the higher the emotionality. By applying this idea to communication, the term will be used to describe the set

of responses to a certain piece of CSR communication and as such making it appear low or high in emotionality. To distinguish the two concepts – emotionality as a characteristic of a person and emotionality of communication content – the emotionality of communication content will be termed as “emotional appeal” which reaches out to the recipient, in this case the consumer.

When studying emotions, what was standing out for many years was the overrepresentation of negative emotions. For a very long time psychological research was primarily focused on negative CAD (contempt, anger, disgust) emotion, in contrast to the evaluation of positive emotional reaction (Xie, Bagozzi and Grønhaug 2015: 339). However, it is argued that this might have been due to earlier research outcomes, that argued that there exist more negative emotions than positive ones (e.g. Shaver et al. 1987; Ellsworth and Smith 1988). Corresponding research outcomes have been made in CSR research, where an asymmetry in the effects of CSR information towards company evaluation was noticed: consumers are more sensitive to negative information than they are to positive one (Sen and Bhattacharya 2001).

However, in recent years this one-dimensional focus on negative emotions became shifted away towards the academic examination of positive emotions, beauty and aesthetics of consumption. Especially the field of psychology investigates the effects of aesthetics and emotions on modern everyday consumption (Venkatesh and Meamber 2008) and it was shown that aesthetics and the positive emotions related to that, influences identity construction, product and brand choices. Especially in economics and public policy research the emotion happiness is of great academic interest. The so-called “happiness research” investigates which conditions feature peoples’ happiness and emotional well-being (Frey and Stutzer 2005; Frey, Stutzer, and Benz 2008; Frey, Gallus, and Steiner 2014).

Though deriving more and more attention, studies on the effects of emotions on human decision- making beyond the field of psychology are rare. In business studies emotions have been studied in relation to consumer responses (Gardner 2015), especially in advertising research (Holbrook and Batra 2015). In the specific field of CSR communication, however, only few prior research on the role of emotions has been carried out. Two mayor exceptions are the recent studies by Pérez and Rodríguez del Bosque 2015) and Xie, Bagozzi, and Grønhaug 2015).

Different from former studies on stakeholder responses (Bhattacharya, Korschun, and Sen 2009) which are strongly based on reasoned processes, Xie, Bagozzi and Grønhaug (2015) argue that

these (reasoned) processes are too complex to grasp and might not be the first response decision- makers may have in moral contexts (p. 334; Haidt 2012). Hence, emotions, in their study “moral emotions” ( see Grappi, Romani, and Bagozzi 2013), might play a very important role in the communication of corporate responsibility. Moral emotions are defined as “combinations of inherited and learned reactions to events violating one’s ethical sensitivities” (Xie, Bagozzi, and Grønhaug 2015). Following Haidt (2012: 45) moral emotions process information on the moral significance of stimuli which one is exposed to. In the respective study the mediating role of the moral emotions gratitude, anger, contempt and disgust was tested with regard to consumer responses to perceived corporate responsive (green) and corporate irresponsive (non-green) actions.

In the model, it is hypothesized that:

1. for non-green actions: the effects of perceived environmental irresponsibility (X) on consumer responses (Y) pass through the moral emotions (M): contempt, anger and disgust, while being governed by the five moderators (W): social justice values, empathy, moral identity, relational self and collective self.

Figure 2: The mediating role of moral emotions in corporate environmental irresponsible actions

Source: Xie, Bagozzi and Grønhaug (2015: 335)

2. for green actions: the effects of perceived environmental responsibility (X) on consumer responses (Y) pass through the moral emotion (M): gratitude, while being governed by the moderator (W): empathy.

Figure 3: The mediating role of moral emotions in corporate environmental responsible actions

Source: Xie, Bagozzi and Grønhaug (2015: 335)

The hypothesized model could be justified in their experimental study, while it was shown that negative emotions have slightly more impact on consumer responses than positive ones (Xie, Bagozzi and Grønhaug 2015: 330).

3.4. Theoretical discussion

After addressing different theories, the theories on credibility, trust and emotions were used to construct a theoretical model for this thesis.

It is hypothesized that the CSR communication with its different emotional appeal (X), provokes distinct emotions for the consumer. These mediated emotions (mE), in turn, influence the perceived credibility and trust by the consumer (Y). The direct influence of CSR communication and its emotional appeal (X) on the perceived credibility and trust by the consumer (Y) is also tested.

Figure 4: Hypothesized model for the quantitative analysis

Mediated emotions (mE)

X: CSR Y: Perceived communication credibility (and trust) H3

With regard to Jackob (Jackob 2008), the constructed model is classified as a recipient model, which allocates credibility to the side of the recipient. It is expected, that when experiencing positive emotions, the consumer will be more likely believe in the credibility of the company (sender) and its messages (content).

Inspired by the theories addressed, the following hypotheses have been constructed:

H1: Different appeal to emotions in CSR communication provokes distinct emotions for the consumer. H2: The perceived credibility (and trust) is influenced by the consumer’s emotions. H3: The different appeal to emotions in CSR communication influences the perceived credibility (and trust).

While the definition for emotions and emotional appeal got clearly addressed, finding a definition for credibility and trust was challenging. As it was stated in the beginning of the theory chapter already, credibility and trust are two very interrelated concepts. When studying the ideas on public trust and credibility, this became apparently clear.

In order to clearly address the issue of credibility and trust, this thesis used Bentele’s (1994; 2008; 2011) definition of public trust and credibility as a sub-concept idea as a starting point, arguing that credibility represents a multiple relationship of communicative content, sender and recipient that is more tangible than the future-oriented trust model.

Thus, it is mainly the issue of credibility (as a sub-concept of trust), which will be addressed. For this the extended understanding of Lock (2016), who adds a normative and communicative dimension to the credibility concept, will be used, though being inspired by the theories on public trust.

When dealing with both theories, public trust and credibility, it became clear that conceptually they are not as far away from each other, as one would think. The social-normative dimension (Bentele and Nothhaft 2011) and the normative appropriateness (Lock 2016) both imply ethical and normative behavior and a sense of responsibility. The social-psychological dimension from public trust assesses the social behavior, character and communicative behavior. This is very similar to what is assessed when approaching sincerity. The understandability of the content shows the company’s competence in terms of successfully communicating with its consumers, thus adding to their professional competence.

Acknowledging, the interconnectedness of trust and credibility, both theories were used for operationalization (see 4.4.3). Nevertheless, it was decided to put a clear focus on the credibility concept by Lock (2016) when measuring the consumers’ responses in the data collection process. However, still seeing credibility as a sub-concept of trust, inferences to trust will be made wherever possible.

More discussion on the theories follow in the method part (4.4.2. and 4.4.3) when operationalizing of the theories is conducted for the data collection process.

4. Methodology

For the research process the decision on the most appropriate method is one of the main challenges. In this respect, it is crucial to put a clear articulation of the research objectives first and only then to decide on the most appropriate method or methods for the data collection (Hansen and Machin 2013).

Following this thought, the research strategy will be outlined and justified first, before the data collection process and further research implications will be presented and explored.

4.1. Research design

The aim of this paper is it to identify the reaction of consumers towards CSR communication with different emotional appeal in respect to the perceived trust and credibility. The target of the applied method are consumers. As a logical consequence, this means that the method needs to address a huge population. Being confronted with a large amount of data quantitative methods are to be preferred.

Above that, the chosen method should enable the visual presentation of different pieces of CSR communication to the participant. Thus, two options have been considered: a laboratory experiment and a web survey with experimental group division. Since a laboratory experiment does not provide natural surroundings, which is especially important when asking the respondents about their emotions, it was decided to carry out an online survey with two groups of participants. One group has been exposed to CSR communication with a low level of emotional appeal, while the second group has been exposed to CSR communication content with a high emotional appeal.

The literature on surveys is very developed. In the media and communication realm, surveys are often used to measure media audiences (Hansen and Machin 2013: 204-226). Though being slightly different, in a broad sense this study also measure the audience (consumers) of a certain media (corporate communication).

The analysis is based on a case study of an international German trading company, which is actively communicating CSR through different channels and with different levels of emotional appeal.

While the survey is the primary source of data and subsequent analysis, background interviews with experts in the company were carried out to contribute to a reasonable research set-up. Above that, it was necessary to identify pieces of CSR communication by the company, which fit into the research frame (see 4.2.)

For the data analysis, it was decided to carry out a statistical analysis that includes four steps. First, the data will be tested for group differences, assessing H1 and H3. After that, the bivariate correlation between credibility/trust and the measured emotions will be assessed. Thereafter, factor analysis is used for data reduction and the extracted factors are then used for the construction of two multiple regression models, one for the text-image content, one for the moving image content.

4.2. Case Study approach

To diminish unrelated influence provoked by inter-company differences in communication strategies and implementation, and the danger of prior brand knowledge, it was decided not only to exclude the brand name from the communication content, but to downscale the scope of the study to one company and its several pieces of CSR communication only. The study, thus, follows a case study approach.

The company investigated is a German trading company from Hamburg, which has a strong brand reputation for coffee and apparel.

To compare the influence of different media, the study investigated both text-image combinations and moving images. Each of the selected communication content was classified as either having a high or low appeal to emotion. A high appeal does not mean that the communication is only targeted to emotions, and a low appeal does not mean that emotions do not play a role at all. It was rather the emphasis and scope of emotions that had to be evaluated to classify the content for the further operationalization.

The classification was done by assessing the communication’s overall focus, selected images, message and language use, as well as sound in the case of moving images. The criteria have been developed by the researcher after carrying out background interviews with representatives of the firm, to avoid subjective influence on the made classification.

Table 5: Criteria for assessment of emotional appeal (own compilation)

Criteria High Low Focus people and their story company achievement/ engagement Images body language and facial show nature, work expression environment, tools etc. Message evaluative informative Language use personal impersonal/ neutral Sound/ Music is used to underline the none, or background music message only

Since the company has several CSR communication initiatives, such as coffee, furniture and textile, which are quite distinct from each other, one branch had to be chosen. The focus of the CSR communication content under research is, thus, responsible action in the apparel sector, particularly the responsible usage of cotton and the engagement for a sustainable supply chain.

It was attempted to operationalize the choice of the communication content and avoid the influence of subjective evaluation. However, especially when investigating emotions being objective is challenging. The researcher acknowledges this challenge and is thus double-checking the evaluation by asking the respondents to classify and rate the emotions they felt while being exposed to the content, before assessing the perceived credibility and trust.

The selected and classified CSR communication is presented in the following table. Appendix 2 presents some exemplary images from the selected CSR content.

Table 6: Classified CSR communication content (own compilation)

High emotional appeal Low emotional appeal Text-image combination GUT GEMACHT – Das WE Nachhaltigkeitsbericht 2014 Programm – Nachhaltige Lieferketten

GUT GEMACHT – Nachhaltigkeitsbericht 2014 Nachhaltige Textilien – Nachhaltige Ressourcen und Produkte Moving image Mary oder Hannah- Wer Nachhaltigkeit – erzählt von ihrem Schultag Gemeinsam mit Cotton made in Africa für die Zukunft in Afrika

Nachhaltigkeit - Das WE Nachhaltigkeit - Wir, das Programm in Vietnam heißt gemeinsam handeln! See Appendix 2 for more details on the selected CSR content

4.3. Data collection

The data collection has been accomplished by an online survey which was set-up with the non- commercial web tool SoSci Survey4. The survey was carried out in German, since the population are German consumers and the selected CSR content is presented in German. Participants have been recruited on survey platforms, via a corporate blog which is dealing with sustainability on a regularly basis and via email.5 By chance the participants were either assigned to group1 (G1) or group2 (G2). G1 was exposed to one text and one video with a low emotional appeal, while G2 was seeing a text and a video with high emotional appeal (with text and video being randomly selected from the classified CSR content in Table 6). Regardless the group, all participants were asked to rate the emotions they felt after reading the text and after watching the video (see 4.3.2).

4 https://www.soscisurvey.de/index.php?page=home&l=eng 5 Consequently, this sample does not fulfil the requirements for a simple random sample (SRS) which is usually one assumption for most statistical tests. The researcher is aware of this shortcoming, however, for the statistical tests SRS will be assumed when necessary.

After classifying their felt emotions, the participants were furthermore asked to answer questions to analyze their perceived credibility of the communication content and the sender (see 4.3.3). Measurement of both emotions and credibility was made on a 5-level Likert-type scale. With regard to Rohrmann (1978) the following linguistic features are considered to be equally distributed across the 5-level scale (in German): “gar nicht – kaum – mittelmäßig – ziemlich – außerordentlich.” and, thus, were used for the assessment. In addition to this, the survey was asking for socio-demographic features (such as age, gender and education) consumer behavior (“How often do you purchase new products, except for food?”) and a short self-assessment on emotionality (“I regard myself an emotional person”, “Feelings guide me through my everyday life” and “I evaluate me and others with respect to moral and normative values and standards.”). In total, this procedure resulted in 71 variables (see Appendix 4 for a complete list of variables).

Before going into the field with the constructed survey, the questionnaire has been tested in a pre- test with 5 individuals. With regard to Hansen and Machin (2013), during this process it was made sure that a) the questions are understood as intended, b) respondents know how to work their way through the questionnaire, i.e. that the respondents are introduced to the procedure and guided well through the questions, c) respondents can ask questions, d) question are worded such that they will yield usable and appropriate answers, i.e. that the questions are standardized/ well operationalized, e) answers are not influenced by particular phrases or word choice. In addition to this, the technical functioning especially of the video content and a coherent and appealing design was made sure of (Couper 2008). Appendix 5 presents some exemplary extracts from the questionnaire.

The online survey was accessible online between 4th March and 9th April 2017 and 103 interviews were recorded. After extracting uncomplete data sets and deleting one that was regarded as invalid due to same response values on all scale measures, 82 valid cases remained for the data analysis. This yields in a completion rate of 80%, which suggests that the survey has been constructed in a comprehensible way.

4.4. Measuring emotions

Following the component definition of emotions (see 3.2.) the ideal measurement method for emotions would need to measure 1) the continuous changes in appraisal process in the nervous system 2) the responses generated through the changes in the nervous system 3) the motivational

changes and action tendencies 4) the patterns of facial and vocal expression and 5) the subjectively experienced feeling state that reflects all of those changes (Scherer 2005: 709). However, due to its complexity such a comprehensive measurement has never been performed and is unlikely to become standard procedure (Scherer 2005: 709).

In the restricted (online) environment of this study, neither cognitive, neurological nor facial or vocal expression could be measured. What could be measured however, was the subjectively experienced feeling state, which has been considered to reflect all those changes. One problem of the measurement of emotions via self-classification by the respondent is the standardization of emotions. One the one hand the standardization might constrain the respondent when feeling emotions that are not present in the measurement scale, on the other hand free response measurement might lead to reduced comparability between groups and individuals. To avoid constraints resulting from standardization, but keeping its advantage, the Geneva Emotion Wheel (GEW; see Scherer 2005) was used to measure the consumers’ emotional responses.

Figure 5: The Geneva Emotion Wheel (GEW) by Scherer (2005)

The GEW has its foundations in the Geneva Affect Label coder, a free response measurement of emotional feeling, which clusters verbal expressions of emotions (such as happy or delighted) in affect categories (such as happiness). As a result, the GEW assesses the quality of emotion in a

two-dimensional space (negative-positive, high control-low control), while at the same time measuring the intensity of emotions (as distance from origin) in their emotional families. The emotional families represent the generic term for emotions that might be named differently, when they are felt in different intensity. For example, irritation might be a less intense emotion belonging to the emotional family of anger, while anxiety is a less intense form of fear. The applied GEW model with its original instructions is to be found in the appendix (see Appendix 3 and 4).

To measure the participants’ emotions in the survey the GEW was operationalized by presenting Figure 5, explaining its features and giving a short instruction of how emotions should be rated. In the following the participants were asked to rate in which intensity they felt the 20 emotional families given: interest, amusement, pride, joy, pleasure, contentment, love, admiration, relief, sadness, guilt, regret, shame, disappointment, fear, disgust, contempt, hate and anger on a 5-level Likert-type scale.

4.5. Measuring credibility and trust

Understanding credibility as a sub-concept of trust and highly intercorrelated with definitions of trustworthiness in particular, it was decided to make use of the concept of public trust (Bentele and Nothhaft 2011) and the theory framework by Lock (2016), which originally was constructed to measure credibility as a separate concept.

Since the future-oriented trust is understood as non-restricted to communicative content, the criteria for credibility could be more easily operationalized for the measurement. However, the dimensions for credibility as defined by Lock (2016) and by Bentele and Nothhaft (2011) are not that different when approaching them in more detail (also see 3.4. on this).

What Lock (2016) approaches with sincerity is closely related to the social-psychological dimension from public trust, which assesses social and communicative behavior and character. Truth plays a role in both concepts, investigating if the consumers believe in the message and the communication of the company. On the other hand, the social-normative dimension (Bentele and Nothhaft 2011) and the normative appropriateness (Lock 2016) both imply ethical and normative behavior and a sense of responsibility. Though the expert dimension gets not visible in the credibility concept of Lock (2016) on first sight, it might be connected to the understandability, showing how competent the firm is an effectively communicating to its consumers.

Summing up, the theories on public trust and credibility have been operationalized using the theoretical concept on credibility by Lock (2016), though stressing the parallels to public trust. Following the model proposed by Lock (2016) the four dimensions Understandability/ Comprehensibility, Truth, Sincerity and Normative Appropriateness are approached.

Table 7: Factors and measurement variables for credibility and trust

Factors Measurement variables Understandability/ Comprehensibility Compr1: I understand the content well. Compr2: The information given is comprehensible.

Truth Truth1: I believe what was stated. Truth 2: I think what was communicated is true.

Sincerity (Social-psychological dimension) Auth1: I believe that the company is authentic. Auth2: I think the communication is genuine.

Normative Appropriateness (Social- NormCor1: I think the company acts morally normative dimension) right NormCor2: I believe the company keeps to moral standards

The table above matches the theory inspired dimensions to the measurement variables. Each dimension is represented by two variables, measured on a Likert-type scale, asking to rate the level of agreement on a scale ranging from 1 (not at all) to 5 (very much). All items together were then used to form an eight-item Likert scale for perceived credibility.

Since the operationalized model for perceived credibility and trust leans more towards credibility in the analysis, this thesis will stick to the term credibility from now on.

4.6. Statistical considerations

Before diving into the data analysis some statistical considerations have been made. This concerns population and sampling, measurement of emotions and credibility and trust, as well as concerns of reliability and validity.

4.6.1. Population and sampling

The population of the survey are German consumers. Considerations have been made to limit the age group to the “advertising-relevant target group”, an age group from 14-49 years which is used in marketing research. However, since ongoing discussions of demographic and societal change in media consumption this limitation is widely criticized for excluding a major part of consumers, that are very relevant for media consumption and possess a great purchasing power (Schwegler 2013). Therefore, this thesis has used the extended “advertising-relevant target group” ranging from 14-59 years. This age group is considered to be most active and engaged with media content and consumption.

4.6.2. Reliability and validity

Validity is understood as the extent to which a measure or a set of measures correctly represents the concept of the study. (Hair et al. 2014: 3; Bryman 2016: 158-162). In case of this thesis, it deals with the question if the measurements for emotions and credibility, actually measure the concepts behind. Validity can be strengthened by applying standard measurements, such as the GEW for emotions, and/or by a strong theoretical background, on that the measurement assumptions are based on (see chapter 3 for theoretical background).

While validity is about what to measure, reliability is about how to measure. Reliability assesses to which extend the measurements are consistent in what it is intended to measure. (Hair et al. 2014: 3; Bryman 2016: 156-158)

The Cronbach’s alpha test is widely employed to measure the internal reliability of a measurement scale. The computed alpha coefficient will vary between 0 (no internal reliability) and 1 (perfect internal validity), with a figure above 0.80 denoting an acceptable level of internal reliability (Bryman 2016). The Cronbach’s alpha test was computed for the measurement of credibility with Compr1, Compr2, Truth1 Truth 2, Auth1, Auth2 NormCor1 and NormCor2 as measurement

indicators and the measurement of Emotions with the 20 different emotions from the GEW measurement scale.

For text-related credibility the test resulted in a Cronbach’s alpha coefficient of 0.891, while the video-related credibility achieved an even higher score of 0.915.

Table 8: Cronbach's Alpha for text-related credibility

Cronbach's Alpha N of Items ,891 8

Table 9: Cronbach's Alpha for video-related credibility

Cronbach's Alpha N of Items ,915 8 The above results account for a high internal reliability of the measurement employed for credibility.

For the measurement of emotions, a widely used standard measurement model, the Geneva Emotion Wheel (GEW) was used, so that reliability was expected to be high. The Cronbach’s Alpha approved that yielding in a coefficient of 0.895 for the measurement of emotions after reading the text and in a coefficient of 0.891 for the measurement after seeing the video.

4.7. Statistical tests

The collected data will be analyzed with the help of different statistical techniques and the Software Package SPSS Statistics to eventually answer the two sub-research questions:

RQ 1.2 Do different forms of CSR content mediate different emotions? RQ 1.3 To which extend do the consumers’ emotions influence the perceived credibility of a company? by assessing the corresponding hypotheses:

For a better understanding the applied statistical tests will be briefly explained.

4.7.1. Hypothesis testing and significance level

For all statistical tests, hypothesis testing is crucial. In general, the hypothesis testing includes the following steps (Gravetter and Wallnau 2009):

1) Translation of the topic of interest to statistical terms by formulating the null hypothesis (and the corresponding alternative hypothesis) that reflects the parameter of interest 2) Setting the significance level α for the test

The significance level is the probability of rejecting the null hypothesis (H0), given that it is true. The significance level is chosen by the researcher with regard to the implied risk of incorrectly rejecting a true hypothesis. The significance level is often set to 0.05%, given that it is acceptable to have a 5% probability of rejecting a true null hypothesis. 3) A random sample is taken and relevant information from this sample is used in the following steps. 4) Determining the appropriate test statistic and calculating it. 5) Calculating the P‐value based upon the value of the test statistic. The P-value is defined as the probability of getting the result which was calculated, or an even more extreme result, given that the null hypothesis is true. If the P-Value is smaller than the significance level, the null hypothesis is rejected.

After carrying out all the relevant steps, the null hypothesis is either rejected or it cannot be rejected from the provided data and thus, still must be regarded as true.

At the end of a hypothesis test, a proper conclusion based on statistical criteria can be formulated.

4.7.2. Likert type items and Likert scale

Likert (1932) proposed a summated scale for the assessment of survey respondent’s attitudes and until today the multi-item Likert scale is widely used in survey research especially in the social sciences. A Likert item in its original sense, was a statement of an attitude, the respondent needs to evaluate using a five-point scale. As such Likert items result in ordinal data, not in numeric data.

By measuring one concept with several Likert items, however, a Likert scale can be composed by creating a summated scale variable, the Likert scale. By using a sufficient number of items, the computed Likert scale thereafter comes close to a scale measure and might be treated like this in statistical analyses (Rasmussen 1989).

4.7.3. Comparing groups with statistical tests

Both, the independent sample t-test and the Mann-Whitney U-Test, compare the values of two different, independent samples or groups. While the t-test is used for metric (scale or interval) data, the Mann-Whitney U-Test is a non-parametric test. (Gravetter and Wallnau 2009)

In general, parametric test possess more power and can detect differences more easily. However, for non-metric data and for data that does not meet the assumption of the t-test (normality, equal variances etc.), the Mann-Whitney U-Test is the better choice.

For the Mann-Whitney U-Test, all observations are ranked and the sums of the ranks for the two groups are compared then. Different from the t-test, which compares the groups by their mean, the Mann-Whitney U-Test works with distribution and the median.

Beyond the advantage of handling ordinal data, the Mann-Whitney U-Test can also handle small group sizes and is more robust to extreme values (outlier).

The null and alternative hypotheses can be formulated in the following ways:

H0: The two distributions are the same

Ha: One distribution has values that are systematically different (systematically larger or smaller)

Or based on the median:

H0: m1 = m2 (the two population medians are equal)

Ha: m1 ≠ m2 (the two population medians are not equal)

The following assumptions need to be fulfilled:

• Independent random sample • Independent groups • The variable can at least be ordered or ranked (ordinal data)

For assumption based on the median, one more assumption needs to be checked. That is the distribution of the variable, which needs to have the same shape in both population groups. After checking the assumptions, the test statistics is to be analyzed, searching for significant results, which then approve that group differences exist.

4.7.4. Bivariate Correlation

The relationship between two variables can be expressed by bivariate correlation. Two correlation analysis are widely used: Pearson correlation and Spearman correlation. (Gravetter and Wallnau 2009)

The Spearman correlation requires variables that can be ranked (ordinal data) and measures the strength and direction of monotonic association between two variables. Different from the Pearson correlation the relationship has not to be linear – but can have any other shape (though including linear relationships).

The Spearman coefficient r can take values from -1 to +1, with -1 indicating a perfect negative relationship, 0 indicating no relationship and +1 indication a perfect positive relationship.

The general assessment of the strength of the relationship looks like the following:

.1 | r | .3 weak .3 | r | .5 medium .5 | r | .. strong

The test hypotheses for correlation were formulated in the following way:

H0: There is no relationship between the variables (ρ = 0)

Ha: There is a monotonic relationship between the variables (ρ ≠ 0)

The standard significance level of 0.05 is used, when approaching the respective correlation. Only significant results account for a relationship between the two variables.

4.7.5. Factor Analysis

Factor analysis is a statistical approach that can be used to explore a large data set for interrelationships among the variables. It can be used to explore the data for underlying dimensions (factors) and for data reduction. Using factor analysis for data reduction, it is the aim to condense the information in the data set into a smaller set of variables with a minimal loss of information (Norman and Streiner 2003: 144; Hair et al. 2014: 16, ch. 3)

The basic assumption for factor analysis is that the correlation among the variables can be explained by underlying factors. By determining the underlying factors in a data set and extracting the most representative, the data set can be reduced without losing information. Therefore, the first step is the evaluation of the KMO and Bartlett’s Test, which indicates if at least some of the variables have significant correlation. A KMO above 0.5 indicates that a factor analysis can be successfully computed, given that the result is significant (p-value < 0.05).

The tests for factor analysis seek for variables that strongly correlated with each other, and at the same time show a weak correlation with another set of variables (Norman and Streiner 2003: 147). For this a correlation matrix is constructed. The decision on how many factors to keep is usually made based on the eigenvalue (Kaiser-Guttman criterion) or by assessing the scree plot. To be kept in the analysis, a factor needs to account for the variance in (at least) one variable, which than results in an eigenvalue of one. Thus, all factor with an eigenvalue of 1 and above are kept in the analysis. The screeplot is a graphical expression with the factors on the x-axis and the eigenvalue on the y-axis. The number of factors is decided based on where the steep part of the line ends and change in variance bottoms out. (Norman and Streiner 2003: 147-148)

After deciding on how many factors to keep, the researcher needs to address the factor loadings to decide on the variables that express the factor in the best manner. The weights of the variables for each factor are shown in the factor-loading matrix. Since there is no consensus about when a factor

loading is significant, in this thesis the equation by Norman and Streiner (2003:149) will be used: 5.152/ √N-2, which results in a minimum loading of 0.58 for the acquired data set.

Another important step in the analysis is rotation. A rotated model tends to give more precise factor loadings by avoiding three main problems: putting too much emphasis on the first (often general) factor, presenting bipolar factors (positive and negative loadings) and factor complexity which is variables loading on several factors. Varimax Rotation is the most widely used rotation method, since it maximizes the variance explained by each factor. (Norman and Streiner 2003:150-153).

Using factor analysis for data reduction, each factor will be treated as a type of scale, where all variables with significant loading for the respective factor will be transformed in a summated scale variable for further analysis (Norman and Streiner 2003:150-154).

4.7.6. Multiple Regression

Multiple linear regression, often just multiple regression, is the appropriate method of analysis when measuring the interactions between several independent variables to predict the change in one metric dependent variable. (Hair et al. 2014)

Multiple regression analysis is by far the most widely used dependence technique, especially in business decision making (Hair et al. 2014). Its biggest advantage is probably, that it can be used for all types of problems ranging from very general to very specific, by relating one or several factors to one specific outcome. By computing parameters, it is also the foundation for forecasting models (Hair et al. 2014).

The equation for a multiple regression analysis might look the following:

Y (Dependent variable) = b0 + b1X1 + b2X2 + … +bnXn

The statistical analysis is conducted by estimating values of the parameters b1 → bn to minimize the squared differences between the real data and the estimated data (statistical rule of least squares). The computed parameters express the change the independent variable has on the dependent variable after all the effects of the other independent variables have been accounted for. (Norman and Streiner 2003).

There exist several assumptions that should be fulfilled for multiple regression:

1. The true relationship is linear 2. The error terms are random variables with mean 0 and constant variance 3. The error terms are not correlated with each other 4. The error terms are normally distributed

The assumption can be easily tested by assessing the partial regression plots for any non-random residuals (for 1-3) and the normal Q-Q plot (4) for any non-normally distributed shape.

The estimated regression model is than to be tested for the overall model fit. The prediction accuracy of the multiple regression model can be evaluated by the coefficient of determination (adjusted R2) and the average distance from the regression line, the standard error of the regression prediction (SEE).

The following hypotheses are used to test if a significant amount of variation is explained by the regression model:

2 H0: R = 0 No part of the variation in Y is explained by the X variables 2 Ha: R > 0 Some part in the variation of Y is explained by the change in X

For the SEE it pertains, the smaller the SEE, the better the model fit.

After assessing the overall model fit, the statistical significance of the model has to be evaluated (Hair et al. 2014). The F ratio tests if the amount of variation explained by the model is more than base line prediction. That is, if R2 is significantly different from zero. The calculated F-value shows the amount of variation explained by the model compared to how much is explained using the simple mean of Y, given that it is significant.

Once the overall significance was assessed, the regression coefficients need to be addressed for significant results. Significant coefficients can be deployed to the regression model and express the impact the variate has on the dependent variables. After interpreting the results, changes in the hypothesized model might be made and the analysis might be repeated with a revised model.

4.8. Ethics

When considering ethical issues involved in the conducted research, it became clear that it is not only ethical guidelines in Sweden that must be addressed, but also the ethical guidelines in the country where the research has been conducted.

In general, it is the ethic commission of the respective research institute or university that deals with ethical concerns and publishes guidelines. However, there is one main exception that is of interest for this study: the professional ethic guidelines for psychologists by the German Psychological Society (Deutsche Gesellschaft für Psychologie e.V. 2016). Since this thesis deals with emotions, these guidelines need to be addressed. The relevant issues that need to be addressed for the present research set-up are:

1. Ensuring a good academic practice (providing documentation on the data, methods, results and references to the public) 2. The need for a voluntary participation that provides clear information on the research 3. Provide the opportunity for detailed information on the research outcomes 4. Ensure that data is collected anonymously.

This thesis follows academic standards and guidelines such as referencing and clear documentation. The data collection was conducted through an online survey. Participation was voluntary and people participating were informed that their data would be used for further analysis regarding this Master thesis. Above that, they were given the possibility to reach out to the researcher via email to resolve questions and to be informed about the research outcomes. The data gathered was automatically anonymized and no inference can be made to individual persons. Thus, the ethical demands to this study have been met.

4.9. Reflection

The aim of the research was it to measure the influence of different CSR communication to the consumer. For this emotional appeal, emotions and perceived credibility have been under investigation. The choice of methods was made after contemplating on the research goals. Since consumers are the targeted research subject, an online survey seemed to be the most effective way in terms of reach and availability.

When deciding on the CSR content, that was going to be used, authentic CSR content was favorized by the researcher. This had one main reason: Authentic CSR content provides a more authentic picture of the “real” CSR communication out there, than constructed content could ever do. However, to ensure comparability among the different participants, it was decided on a case study approach. Using authentic CSR content from one company, is considered as the best trade-off between staying close to reality, whereas ensuring comparability.

When planning the data collection process, it became clear, that for the scope of this research project, a simple random sampling (SRS) was not realizable. However, to ensure a wide-range of participation, the survey has been incentivized and promoted on different portals (see 4.3.) to generate a sufficient amount of data for the statistical analysis. Above that, socio-demographic features have been collected to test the data for diversity and representation of the target population.

Though a sufficient minimum amount of answers could be collected with both groups including more than 30 respondents, sample size is still an issue, that might be the target of criticism. The conclusion that are drawn based on this sample, thus, need to be handled with care and should be reviewed using a wider sampling.

Regarding the data analysis, several statistical tests have been compiled to ensure well-grounded conclusions. The data have been checked for the respective test assumptions before the test statistics have been calculated. Since the variables have been measured using Likert type items on an ordinal scale, test statistics for ordinal data have been preferred whenever possible. Though being widely debated Likert scales, that include a sufficient amount of Likert items, might be treated as continuous variables under certain circumstances. For the multiple regression model this assumption was made use of for the summated scale variable credibility.

Moreover, reliability and validity of the data have been addressed by using a strong theory background and internally consistent measurement variables. For both, emotions and credibility, good test statistics indicate that the demands for reliability and validity of the data have been met in this study.

5. Results

In a first step, the data have been analyzed with regard to their sociodemographic features, in order to be aware of potentially biased data resulting from the participants’ features such as age, gender and occupation. Since this study investigates consumers, a review of consumption behavior was included. The age distribution in the sample is skewed towards the lower age groups with the majority of the participants being aged 20 – 34 years (Appendix 6). This might be explained by a more active usage of the internet. However, the data provide answers from all age groups under investigation (18-59 years), so that unintended influence resulting from age is diminished. Regarding the sociodemographic variable occupation (Appendix 6), the survey was completed by different groups including pupils, civil servants and job seekers. However, the majority (N=37) are students, closely followed by employees (N=28), which accounts for a more educated sample than average population. The majority of the respondents consume products (others than food products) several times a month or more (N=52), so that it can be concluded that we have active consumers present in our sample. For further analysis, the participants have been assigned to experimental groups. 39 respondents have been assigned to G1 and 43 respondents to G2.

5.1. Testing group differences

In a first step, the experimental groups G1 (low emotional appeal) and G2 (high emotional appeal) have been compared for significant differences in mediated emotions and perceived credibility.

5.1.1. Differences in mediated emotions

To test H1, the two experimental groups G1 (low emotional appeal) and G2 (high emotional appeal) have been compared to check if a significant difference between the emotions evaluated by the participants exists.

The testing hypotheses have been stated as the following:

H0: The distributions for G1 and G2 are the same.

Ha: One distribution has values that are systematically different (systematically larger or smaller)

The significance level was set to 0.05.

The appropriate test to measure differences in two groups with ordinal data is the Mann-Whitney U-Test, which therefore has been performed in the following (Appendix 8).

Since the respondents were asked to rate their emotions twice, after reading a text (text-related emotions) and after watching a video (video-related emotions), the group comparison has been carried out twice.

For the text-related emotions a significant difference for Guilt has been detected. The distribution is significantly larger in G1 (see Appendix 8).

Table 10: Mann-Whitney U-Test Output for Guilt (text-related)

Asymp. Sig. (2- Mann-Whitney U Wilcoxon W Z tailed) Emotion: Guilt 603,000 1549,000 -2,288 ,022

For video related emotions, a significant difference for Regret could be observed. Like Guilt, the distribution for G1 is significantly larger compared to G2 (see Appendix 8).

Table 11: Mann-Whitney U-Test Output for Regret (video-related)

Asymp. Sig. (2- Mann-Whitney U Wilcoxon W Z tailed) Emotion: Regret 643,500 1589,500 -2,078 ,038

For all other emotions, no significant difference between the groups has been detected. Consequently, the null hypothesis of equal distribution in the groups cannot be rejected, with exception for Guilt (text-related) and Regret (video-related).

The boxplots below, visualize the different distribution of Guilt and Regret in G1 and G2. 50% of the respondents (visualized by the green box) in G2 rated the intensity of feeling Guilt below the mean of intensity in G1. For Regret the distribution looks more evenly distributed, though the means are quite different.

Figure 6: Boxplots for Guilt and Regret

5.1.2. Differences in perceived credibility

To assess H3 the variables Compr1, Compr2, Truth1 Truth 2, Auth1, Auth2 NormCor1 and NormCor2, which all together test the amount of perceived credibility and trust, have been tested for significant differences in G1 (low emotional appeal) and G2 (high emotional appeal). For the data analysis, the Mann-Whitney U-Test (Appendix 8) has been used and it was hypothesized that:

H0: The distributions for G1 and G2 are the same.

Ha: One distribution has values that are systematically different (systematically larger or smaller)

The significance level has been set to the standard level of 0.05 since no negative consequences for people have been expected. The assumptions were fulfilled.

Just as in 5.1.1 the Whitney-Mann U-Test was carried out. First for text-related credibility and then for video-related credibility.

For the text-related credibility, a significant difference for Compr2 was detected. The distribution for G1 is significantly smaller compared to G2.

Table 12: Significant difference between G1 and G2 for credibility and trust (text-related)

Asymp. Sig. (2- Mann-Whitney U Wilcoxon W Z tailed) Compr2: The information 590,000 1370,000 -2,521 ,012 given is comprehensible.

For video-related credibility, the test resulted in significant difference for Auth2, which is significantly lower in G2.

Table 13: Significant difference between G1 and G2 for credibility and trust (video-related)

Asymp. Sig. (2- Mann-Whitney U Wilcoxon W Z tailed) Auth2: I think the 629,000 1575,000 -2,023 ,043 communication is genuine.

For all other credibility and trust variables, no significant difference between the groups could be detected (see Appendix 8) and the null hypothesis of equal distribution cannot be rejected.

In the following, additional statistical tests have been carried out, to see how the measured emotions correlate to the measured credibility.

5.2. Correlation between emotions and credibility

After comparing the two test groups for emotions and credibility, the bivariate correlation between measured emotions and the credibility items has been tested directly, to further assess RQ1.3 (see Appendix 9).

For this a Likert scale has been computed by forming a summated scale of the measured items for credibility: Compr1, Compr2, Truth1 Truth 2, Auth1, Auth2 NormCor1 and NormCor2.

Since the measured emotions resulted in ordinal data, the Spearman correlation was used for the correlation analysis.

The test hypotheses are the following:

H0: There is no relationship between the variables (ρ = 0)

Ha: There is a monotonic relationship between the variables (ρ ≠ 0)

For text-related credibility the following correlations have been detected (Appendix 9):

Table 14: Strength and direction of the correlation between credibility and individual emotions (text-related)

Correlation Positive Negative Weak Interest Disgust Love Hate Relief

Medium Pride Contempt Joy Anger Contentment Admiration

For video-related credibility the detected correlations overall were slightly stronger and, in addition to that, only one negative correlation exists (Appendix 9):

Table 15: Strength and direction of the correlation between credibility and individual emotions (video-related)

Correlation Positive Negative Weak Pride Contempt Love

Medium Joy Contentment Admiration Relief Compassion

Strong Interest

In addition to the test of correlation for the individual emotions, a Likert scale has been composed for all emotions (T_Emotions and MI_Emotions), for the positive emotions (T_posEmotions, MI_posEmotions) and the negative emotions (T_negEmotions, MI_negEmotions), to test the

correlation between emotion and credibility in a more generalized way. The division between positive and negative emotions has been made in accordance with the GEW measurement model (see 4.2.2).

Though a Likert scale might be treated as a metric variable in some cases, after compiling both Pearson and Spearman, it was decided to use Spearman correlation to be more precise and avoid misinterpretation.

Table 16: Spearman correlation for Credibility and Emotions (text-related)

T_Emotions T_posEmotions T_negEmotions

Spearman's T_Credibility Correlation -,016 ,343* -,240* rho Coefficient Sig. (2-tailed) ,887 ,002 ,030 N 82 82 82 T_posEmotions Correlation ,589* 1,000 ,331* Coefficient Sig. (2-tailed) ,000 . ,002 N 82 82 82 Significant results have been marked with *

For text related credibility, a medium positive correlation could be detected with positive emotions and a weak negative correlation with negative emotions. No significant correlation exists between all summated emotions and credibility. Worth the mention is the medium positive correlation for positive and negative emotions, which shows that there exists an interrelationship between positive and negative emotion for the text-related analysis.

For the video content, the correlation between positive emotions and credibility becomes even clearer. With a Spearman coefficient of .441 there exists a medium positive relationship between the two variables. Different from the text-related analyses, negative emotions showed no significant correlation with credibility. Additionally, the correlation between positive and negative emotions does not exist for the video content.

Table 17: Spearman correlation for Credibility and Emotions (video-related)

MI_Credibility MI_Emotions MI_posEmotions MI_negEmotions Spearman's MI_Credibility Correlation 1,000 ,179 ,441* -,094 rho Coefficient Sig. (2- . ,107 ,000 ,401 tailed) N 82 82 82 82 Significant results have been marked with *

In the next step, the different dimensions of credibility have been approached, which are Understandability, Truth, Sincerity and Normative Appropriateness.

For the text-related measurement positive emotions revealed to have a medium positive influence on Truth, Sincerity and Normative Appropriateness, though having no significant effect on Understandability. Negative emotions, however, show a negative weak to medium influence on Understandability and Normative Appropriateness.

Table 18: Correlation for dimensions of credibility and emotions (text-related)

posEmotions negEmotions Spearman's rho Understandability -,064 -,388*

,570 ,000 82 82 Truth ,311* -,191 ,004 ,085 82 82 Sincerity ,312* -,135 ,004 ,225 82 82 Normative ,386* -,239*

Appropriateness ,000 ,031 82 82 Significant results have been marked with *

For the video-related measurements, positive emotions similarly to the text-related analysis, revealed a medium positive influence on Truth, Sincerity, and Normative Appropriateness.

However, in contrast to that, the negative emotions did not have any significant influence on Normative appropriateness as they had for the text-related credibility. Nevertheless, negative emotions showed a negative medium influence on Understandability.

Table 19: Correlation for dimensions of credibility and emotions (video-related)

Correlations MI_posEmotion MI_negEmotion s s Spearman's rho Understandability ,167 -,381* ,133 ,000

82 82 Truth ,395* ,003 ,000 ,975 82 82 Sincerity ,323* -,067 ,003 ,552 82 82 Normative ,435* -,119 Appropriateness ,000 ,289 82 82 Significant results have been marked with *

After assessing bivariate correlation between the measured variables, underlying factors for emotions have been extracted in the next step.

5.3. Factor analysis

Before using multivariate analysis, which tend to be quite complex, factor analysis has been used to test the variables for their explanation power and intercorrelation.

5.3.1. Factor analysis for data reduction

The main aim of the compiled factor analysis has been a data reduction without losing relevant information from the data.

Data reduction turns up to be especially important for the measurement of emotions, since 20 emotions have been measured in the survey.

For text-related emotions the KMO and Bartletts’s test resulted in a KMO of .820 (p-value = .000) and for video-related emotions in a KMO of .856 (p-value =.000). Thus, factor analysis can be used for both sets of measurements.

Table 20: KMO and Bartlett's Test for text-related emotions

KMO and Bartlett's Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy. ,820 Bartlett's Test of Sphericity Approx. Chi-Square 1116,674

df 190 Sig. ,000

Table 21: KMO and Bartlett's Test for video-related emotions

KMO and Bartlett's Test Kaiser-Meyer-Olkin Measure of Sampling Adequacy. ,856 Bartlett's Test of Sphericity Approx. Chi-Square 1345,273 df 190 Sig. ,000

To assess the number of factors, that should be included into the model, two criteria have been used: the latent root criterion (also known as Eigenvalue) and the scree test criterion.

The table below (Table 22) shows that four factors have an Eigenvalue greater than one and together account for more than 70% of the variance in the model.

Table 22: Factor reduction for text-related emotions

Total Variance Explained Extraction Sums of Squared Rotation Sums of Squared Initial Eigenvalues Loadings Loadings % of Cumulative % of Cumulative % of Cumulative Component Total Variance % Total Variance % Total Variance % 1 7,245 36,224 36,224 7,245 36,224 36,224 4,552 22,762 22,762 2 4,159 20,797 57,022 4,159 20,797 57,022 4,330 21,650 44,411 3 1,667 8,337 65,359 1,667 8,337 65,359 3,848 19,242 63,654 4 1,113 5,564 70,923 1,113 5,564 70,923 1,454 7,269 70,923 5 ,785 3,925 74,847 6 ,697 3,485 78,332 7 ,569 2,843 81,176 8 ,538 2,691 83,867 9 ,515 2,576 86,443 10 ,439 2,196 88,639 11 ,398 1,989 90,628 12 ,345 1,724 92,352 13 ,314 1,569 93,920 14 ,295 1,477 95,398 15 ,259 1,295 96,693 16 ,202 1,009 97,702 17 ,151 ,755 98,457 18 ,125 ,627 99,083 19 ,119 ,593 99,676 20 ,065 ,324 100,000 Extraction Method: Principal Component Analysis.

For video-related emotions the test resulted in a similar component matrix with four factors accounting for more than 75% of the variance (see Table 23 below).

Table 23: Factor reduction for video-related emotions

Total Variance Explained Extraction Sums of Squared Rotation Sums of Squared Initial Eigenvalues Loadings Loadings % of Cumulative % of Cumulative % of Cumulative Component Total Variance % Total Variance % Total Variance % 1 7,170 35,851 35,851 7,170 35,851 35,851 5,732 28,661 28,661 2 5,148 25,740 61,591 5,148 25,740 61,591 5,118 25,588 54,249 3 1,758 8,788 70,379 1,758 8,788 70,379 2,466 12,328 66,577 4 1,010 5,050 75,429 1,010 5,050 75,429 1,770 8,852 75,429 5 ,722 3,612 79,041 6 ,584 2,922 81,962 7 ,503 2,513 84,475 8 ,489 2,445 86,920 9 ,420 2,102 89,022 10 ,390 1,950 90,972 11 ,327 1,634 92,606 12 ,276 1,380 93,986 13 ,233 1,167 95,153 14 ,221 1,107 96,261 15 ,178 ,892 97,153 16 ,171 ,854 98,007 17 ,141 ,705 98,712 18 ,118 ,589 99,301 19 ,095 ,473 99,774 20 ,045 ,226 100,000 Extraction Method: Principal Component Analysis.

In line with the Eigenvalue tests, the scree tests for text-related and video-related emotions accounted for four factors to be included in the model.

Figure 7: Scree plots for factor analysis

By interpreting the component matrix, it has been determined which factors should be used to account for the optimum variance in emotion. To add more certainty to the model, Varimax factor rotation has been used and factor loadings have been compared for the non-rotated and rotated model.

The unrotated and the rotated model, looked quite different for both analyses. However, the rotated model is considered to give a higher accuracy. Hence, in case of doubt the rotated model was consulted.

Interpreting the factor loadings, a factor loading above 0.58 is considered as significant (see 4.5.5) and has been kept as an indicator variable for the respective factor.

The following variables have been extracted as indicator variables for each factor (Appendix 10):

Table 24: Extracted factors and representative variables (text-related)

Factor 1 2 3 4 Pride Sadness Amusement Interest Variables with Joy Guilt Disgust Compassion significant Pleasure Regret Contempt factor loadings Contentment Shame Hate Love Disappointment Anger Admiration

Table 25: Extracted factors and representative variables (video-related)

Factor 1 2 3 4 Shame Pride Sadness Interest Variables with Disappointment Joy Guilt Amusement significant Fear Pleasure Regret Compassion factor loadings Disgust Contentment Contempt Love Hate Admiration Anger Relief

5.3.2. Correlation between extracted factors and dimensions of credibility

As an additional step in the data analysis, the extracted factors have been used to check for correlation with the four dimensions of credibility.

Table 26: Correlation between extracted factors and dimensions of credibility (text-related)

Correlations T_Fact1 T_Fact2 T_Fact3 T_Fact4 Spearman's rho Understandability -,055 -,405* -,407* ,060 ,625 ,000 ,000 ,591 82 82 82 82 Truth ,322* -,135 -,228* ,246* ,003 ,228 ,039 ,026 82 82 82 82 Sincerity ,330* -,051 -,201 ,228*

,002 ,649 ,070 ,039 82 82 82 82 Normative ,392* -,162 -,294* ,263* Appropriateness ,000 ,147 ,007 ,017 82 82 82 82 Significant results have been marked with *

For the text-related dimensions of credibility, Factor 2 and 3 have a negative influence on Understandability. Truth is positively affected by Factor 1 (medium correlation) and Factor 4 (weak correlation), while Factor 3 has a weak negative influence.

Sincerity and Normative Appropriateness are positively affected by Factor 1 and 4. Whereas Sincerity is not negatively affected, Factor 3 has a negative influence on Normative Appropriateness.

Table 27: Correlation between extracted factors and dimensions of credibility (video-related)

Correlations MI_Fact1 MI_Fact2 MI_Fact3 MI_Fact4 Spearman's rho Understandability -,464* ,170 -,269* ,202 ,000 ,127 ,015 ,068 82 82 82 82 Truth -,004 ,367* ,062 ,392* ,969 ,001 ,579 ,000 82 82 82 82 Sincerity -,051 ,281* ,031 ,344*

,651 ,011 ,779 ,002 82 82 82 82 Normative -,120 ,395* ,015 ,404* Appropriateness ,283 ,000 ,894 ,000 82 82 82 82 Significant results have been marked with *

The correlation analysis for video-related dimension of credibility and extracted factors resulted in slightly different results. Whereas Understandability in line with the text-related analysis, is only negatively affected by Factor 1 (medium correlation) and by Factor 3 (weak correlation), for Truth, Sincerity and Normative Appropriateness no negative influence has been registered. Truth is positively influenced by Factor 2 and Factor 4, the same accounts for Sincerity and Normative appropriateness.

5.4. Multiple regression

For the multiple regression model, it was hypothesized that a) For the text-related model:

Y (Credibility/Trust) = b0 + b1X1 (T_Fact1) + b2X2 (T_Fact2) + b3X3 (T_Fact3) +

b4X4 (T_Fact4) with T_Fact1, T_Fact2, T_Fact3, T_Fact4 representing the summated scale for the representative variables of the extracted factors. b) For the Video-related model:

Y (Credibility/Trust) = b0 + b1X1 (MI_Fact1) + b2X2 (MI_Fact2) + b3X3 (MI_Fact3) + b4X4 (MI_Fact4) with MI_Fact1, MI_Fact2, MI_Fact3, MI_Fact4 representing the summated scale for the representative variables of the extracted factors.

Since the dependent variable Y (Credibility/Trust) has been measured with a multi-item Likert scale, multiple regression was used, interpreting the computed Likert scale as a scale variable.

The assumption for random error residuals and normality have been addressed. The respective plots did not show any non-random patterns, so that the assumption for multiple regression have been met.

5.4.1. Regression model for perceived credibility and trust after reading the text

First, the analysis addressed the model fit for the hypothesized model (see above).

The test results have shown that the independent variables in the model account for 30,5% of the variance in the dependent variable for perceived credibility and trust (R2 = .305). The significant F-value (8.432) indicates a model that is 8 times as good than by chance.

However, only two of the factors have shown a significant result in the coefficient analysis (Factor 1 and Factor 3). Since no significant explanation power for Factor 2 and 4 could be proved, both

factors have been deleted from the model, before computing the coefficients for the new regression model.

After the deletion of Factor 2 and 4, the R2 decreased to 28.6%, while at the same time the F-value improved to 15.856. Though the model is explaining slightly less variation in Y, the overall model fit is then nearly twice as good.

Table 28: Computed coefficients for the regression model (text-related)

Unstandardized Coefficients Model B Std. Error Sig.

1 (Constant) 3,303 ,201 ,000 T_Fact1 ,304 ,066 ,000 T_Fact3 -,347 ,092 ,000

The computed equation for the revised text-related regression looks like the following:

Y (Credibility/Trust) = 3.303 + 0.304*X1 (T_Fact1) - 0.347*X3 (T_Fact3)

5.4.2. Regression model for perceived credibility after watching the video

For the text-related regression model, the R2 indicated that the independent variables account for 25% of the overall variation in Y. Above that, the significant F-value accounted for a model that is 6 times as good in predicting Y than it would be by chance.

The coefficient analysis, however, showed significant results exclusively for Factor 2 and 4, so that in a second analysis Factor 1 and 3 have been excluded from the model. Like for the text-related revised model, the R2 was slightly diminished after extracting the two non-significant factors, whereas the F-value increased to 12.315 which accounts for a model that predicts the dependent variable 12 times as good than just by chance.

The table below (Table 29) shows the computed coefficients for the regression model:

Table 29: Computed coefficients for the regression model (video-related)

Unstandardized Standardized Coefficients Coefficients Model B Std. Error Beta t Sig. 1 (Constant) 2,257 ,291 7,749 ,000 MI_Fact2 ,217 ,079 ,292 2,741 ,008 MI_Fact4 ,248 ,090 ,292 2,741 ,008

The equation for the revised regression model for perceived credibility after watching the video content looks the following:

Y (Credibility/Trust) = 2.257 + 0.217*X2 (MI_Fact2) + 0.248*X4 (MI_Fact4)

6. Analysis

The statistical tests revealed that emotions have a significant influence on the perceived credibility by the consumer.

The aim of the statistical analysis was it to answer the two sub-research questions RQ 1.2 Do different forms of CSR content mediate different emotions? and RQ 1.3 To which extend do the consumers’ emotions influence the perceived credibility of a company? by testing the corresponding hypotheses H1: Different appeal to emotions in CSR communication provokes distinct emotions for the consumer, H2: The perceived credibility (and trust) is influenced by the consumer’s emotions and H3: The different appeal to emotions in CSR communication influences the perceived credibility (and trust) and to verify the theory inspired hypothesized model (see 3.4.):

Mediated emotions (mE)

Y: Perceived X: CSR communication credibility (and trust)

For RQ 1.1. a measurement model has been developed before starting the data collection process. The measurement model measures the consumers’ emotions and the perceived credibility of the company for different forms of CSR communication. This model can be employed for any future investigations on the influence of emotions and emotional appeal on perceived credibility and trust in CSR communication.

For RQ 1.2 the test results showed a significant difference for guilt (text-related) and regret (video- related) between the two test groups. It is concluded from the results, that people reading a text

with high emotional appeal feel significantly less guilty than people reading a text with low emotional appeal and consumers watching a video with high emotional appeal feel less regret than people watching a video with low emotional appeal.

Thus, the emotions mediated in different types of CSR content can be significantly different from each other, as it was shown for Guilt and Regret. However, though differences in the distribution for all other emotions could be hypothesized from the data, these differences were not large enough to emerge in significant results. Consequently, H1 is only partly approved. A possible reason for that, is that the selected CSR content, which is authentic and not constructed as it done often in experimental studies, was not diverse enough in terms of emotional appeal.

By testing the group differences for perceived credibility, the direct effect of different emotional appeal in CSR communication on perceived credibility has been tested (H3). Here, a significant difference in comprehensibility could be detected. Emotional text-image content was easier to understand than the low emotional content. However, the emotional video content appeared to be less genuine and is as such rated as less credible in terms of sincerity. For H3 it can be concluded that different emotional appeal influences sincerity and comprehensibility of the CSR content. However, the main aim of this thesis was it to study the effect mediated emotions have on the perceived credibility and trust (see RQ 1.3).

Continuing the analysis, the bivariate correlations between the 20 measured emotions and credibility has been assessed.

Influence of emotions for text-image content

The analysis for the measurement after reading the text, showed significant correlation for 11 emotions. Seven emotions have a positive influence on perceived credibility and trust and all these seven emotions are positive emotions. CSR text-image content is, thus, perceived as more credible and trustworthy, when the consumer feels Interest, Love, Relief (weak correlation) and even more when feeling Pride, Joy, Contentment and Admiration (medium correlation). At the same time, there exist four negative emotions, all four being classified as high control emotions (see 4.4.2), that have a negative influence on perceived credibility and trust for CSR text-image content: Disgust and Hate (weak correlation) and Contempt and Anger (medium correlation).

Consequently, for the text-image content a positive influence on perceived credibility is created by the above mentioned positive emotions, and a negative influence by the above mentioned negative ones, though there exist more positive emotions influencing the perceived credibility of the CSR content, than negative ones. When analyzing the different dimensions of credibility, it was observed that positive and negative emotions influence the dimensions in different ways. Whereas, the dimension of understandability is affected by negative emotions only, which have a medium negative impact (-0.388), truth and sincerity record medium positive influence by positive emotions exclusively. The fourth dimension normative appropriateness, however, is influenced by both positive (+0.386) and negative emotions (-0.239).

It is concluded from this, that positive emotions positively influence truth, sincerity and normative appropriateness of CSR text-image content, though not resulting in an extended understanding of the content. Negative emotions, on the other hand, result in a diminished understanding and a more negatively perceived normative appropriateness of the CSR text-image content, though having no impact on perceived truth and sincerity of the content.

Influence of emotions for moving image content

In a similar analysis of bivariate correlation, the influencing emotions for perceived credibility and trust for CSR video content were examined. Remarkably, negative emotions could be analyzed as having a minimal effect for the credibility of CSR video content. Only Contempt showed to have a weak negative correlation with perceived credibility. At the same time, positive emotions showed to have an even stronger positive influence. Eight positive emotions show positive correlation with perceived credibility. Only two (Pride and Love) showed a weak correlation, all other (Joy, Contentment, Admiration, Relief and Compassion) have a medium influence, while Interest has a strong influence.

Thus, for CSR video content it was shown that negative emotions have nearly no influence on the perceived credibility and that in general positive emotions have an even stronger positive effect on perceived credibility. For CSR video content, it seems very important to create positive feelings for the consumer, for the company to appear credible and trustworthy. But which dimensions of credibility get influenced by the positive and negative emotions? Understandability gets negatively influenced by negative emotions, while truth, sincerity and normative appropriateness are not influenced by negative emotions at all. However, positive emotions show a positive impact on

truth, sincerity and strongest on normative appropriateness, though not enhancing understandability. Interpreting the correlations, positive emotions enhance credibility in terms of truth, sincerity and normative appropriateness. Negative emotions on the other hand, only influence understandability of CSR content in a negative way, while having no significant influence on perceived truth, sincerity or normative appropriateness of the CSR video content.

Interpreting underlying factors

The third step in the analysis, was computing a factor analysis to investigate underlying factors in the measured data and to reduce the data set of 20 measured emotions to a more assessable set of variables with condensed information for the subsequent multiple regression model.

By computing factor analysis, awareness was raised that the collected data set of emotions, consists of four underlying factors, which resulted in similar data structures for text-related and video- related emotions. One factor (Factor 1 for text-related, Factor 2 for video-related) consists of positive emotions, such as Joy, Pride, Pleasure, Contentment. A second factor is represented by negative low-control emotions such as Sadness, Guilt, Regret (Factor 2 for text-related, Factor 3 for video-related), a third (Factor 3 for text-related, Factor 1 for video-related) from negative but high-control emotions such as Anger, Hate, Contempt (surprisingly containing Amusement in the text-related analysis), and the fourth factor being made up by Interest and Compassion (and Amusement for the video-related analysis). What has been detected during the factor analysis is the effect of Amusement, which is the only positive emotion that carried a negative factor loading for both text-related and video-related analyses, and thus is having a negative influence on perceived credibility. Trying to provoke the emotion Amusement in a CSR context, thus, results in a more negative attitude by the consumer towards the firm’s credibility.

Multiple Regression for perceived credibility

Continuing with the multiple regression model, only two factors per analysis (text-related and video-related) showed significant influence on perceived credibility. For the text-related regression model this was Factor 1 with the positive emotions, that resulted in a parameter of 0.304 and thus is having a positive influence on perceived credibility, and Factor 3 with the negative high-control emotions (and amusement) with a parameter of -0.347 and thus having a negative influence on perceived credibility.

For the video-related model, the picture looks quite different. Factor 2 with the positive emotions (parameter 0.217) and Factor 4 with Interest, Amusement and Compassions (parameter 0.248) proved to be significant. Thus, for the video content, only positive influence on credibility is created by emotions. Positive emotions have a positive effect on the perceived credibility, while negative emotions have no significant effect.

Influence of emotions on the dimensions of credibility

When assessing the relationship between the extracted factors and the dimensions of credibility, it evolves for the text-related analysis that Factor 1 representing the positive emotions has a medium positive influence on perceived truth, sincerity and normative appropriateness. Factor 3, representing the negative high-control emotions (and amusement), however, has a medium negative impact on understandability and a weak negative influence on normative appropriateness. Just as for the correlation between credibility dimensions, and positive and negative emotions, the negative effect of negative emotions, especially the high-control ones, on understandability gets visible here. For normative appropriateness this effect gets diminished, because of Factor 1 and 3 having opposing influence on this dimension, while positive and negative emotions have been shown to be intercorrelated.

For the video-related analysis, however, Factor 2 with the positive emotions shows a medium positive influence on truth and normative appropriateness, and a weak positive influence on sincerity, while Factor 4 (represented by Interest, Compassion and Amusement) shows an even stronger influence on truth, sincerity and especially on normative appropriateness. In line with the former analyses, all dimensions, but understandability get positively affected by positive emotions, whereas negative emotions diminish the perceived understandability exclusively.

Analysis Summary

Taken all the test results together, H1 could be partly proved by detecting group differences for Guilt and Regret. For RQ 1.2 it can be stated that different CSR content mediates different emotions to the consumer, though in the given framework only two statistical significant differences could be detected. Above that, it was proven that different CSR communication, in terms of the medium used (text-image and moving image), results in different emotions.

H2 was multiply proved to be correct by applying different statistical tests. For RQ 1.3 it was shown that for the text-related both positive and negative emotions influence the perceived credibility. This could be further justified in the multiple regression model. One factor represented by positive emotions such as Joy and Contentment revealed a positive effect on the dependent variable (Y: credibility), whereas a second factor represented by negative high-control emotions such as Anger and Hate, proved to negatively influence the dependent variable. Regarding the dimension of credibility, it is understandability and to a smaller degree normative appropriateness that get negatively affected by the negative high-control emotions. The positive effects get visible for truth, sincerity and normative appropriateness.

For the video-related emotions, however, negative emotions have no significant influence on the perceived credibility. Besides the factor that is represented by positive emotions, it is the factor represented by interest and compassion that has a remarkable influence on perceived truth, sincerity and normative appropriateness.

7. Conclusion and Discussion

This thesis started by exploring the field of CSR and CSR communication research. Though many studies on the effect of CSR communication exist, they are widely limited to business outcomes such as purchase intention or brand reputation. However, some studies from the social science contribute to the bigger picture by addressing credibility and trust and its role in CSR communication. Nevertheless, emotions and their influence in CSR communication, remain an under-researched area, that researchers from different disciplines only recently became aware of.

The aim of this research project was it to contribute to this evolving research field by addressing the influence of emotions and emotional appeal in authentic CSR communication on perceived credibility and trust. Two different media have been studied: text-image content and moving images of a German trading company. In an online survey (N = 82), German consumers were allocated to G1 (low emotional content) or G2 (high emotional content), both groups being exposed to one text-image CSR content and one moving image on CSR. After this, they were asked to rate the intensity of their felt emotions using the Geneva Emotion Wheel (GEW) and to express their agreement on several Likert items to measure perceived credibility (see chapter 4.4.) This constructed measurement model (RQ 1.1.), is widely applicable for all other forms of CSR communication and can be used for further investigations on the role of emotions and its influence on credibility.

Moreover, different statistical tests have been carried out, to analyze the collected data, including testing for group differences, bivariate correlation, factor analysis and multiple regression. With regard to the proposed research questions, it can be concluded, that different CSR content mediates different emotions by its emotional appeal. This could be justified for the two groups G1 and G2, but holds also true for the different media (text-image and moving image). Furthermore, the emotional appeal of the CSR content influences the perceived credibility in multiple ways. While the high emotional content enhances the comprehensibility, it could be proved that it also diminishes sincerity of the CSR communication and the company behind.

For the text-image content, 11 mediated emotions had an impact on perceived credibility. It was proven that six positive emotions have a positive influence on perceived credibility and trust, whereas Amusement and four negative emotions have a negative impact.

The moving image content showed stronger overall correlation between emotions and perceived credibility. Especially, the emotion Interest has a strong positive relationship with perceived credibility. Moreover, the analysis showed that negative emotions have no significant effect on perceived credibility for the video content. Positive emotions, represented by two factors, on the other hand, have a very positive influence. To the contrary, for text-image content, both positive and negative (high-control) emotions influence perceived credibility, with negative emotions having a slightly stronger effect.

In line with earlier research on emotions in CSR communication (Pérez and Rodríguez del Bosque 2015; Xie, Bagozzi, and Grønhaug 2015), the mediating function of emotions could be proved. Emotions have a strong influence on consumer responses, especially in CSR communication. Different from former research, however, negative emotions proved to have slightly more influence on consumer responses regarding the text-image content only, whereas the moving image content was influenced by positive emotions exclusively. Consequently, successful CSR communication is not about avoiding the creation of negative emotions. It is in contrast more about creating positive emotions, while finding a balance between too much and too little emotional appeal.

7.1. Future research

Studying different CSR communication initiatives and campaigns, their outcomes and perception in the public is a relevant area to study for media and communication studies, whether it is investigating communicative and media conditions, in which companies act in, studying how firms adapt to media logics and social expectations or searching for actors in a public discourse on responsibility issues.

Since this study could prove that emotions have a substantial impact on credibility, theoretical as well as empirical approaches should take emotions more into account when dealing with credibility and trust, but also more generally when researching CSR communication and organizational communication from a social science perspective.

When dealing with credibility and trust concepts, it became clear that a conceptual separation of two concepts remains challenging. For this study, the concept of credibility appeared to be more tangible, since it was mainly about studying communication content. However, the close connection between trust and credibility became visible, and deserves further clarification and

theoretical, as well as empirical examination. Though parallels to trust have been drawn in this study, a follow-up study with a similar research set up testing emotions in CSR communication for the influence on trust in a company would eminently contribute to the understanding of the role of emotions in organizational communication.

Though reliability and validity of the study have been assessed as relatively high by using internally consistent measurements, strong theory and standard measurements, the conclusions have been drawn based on a sample of N=82. Future research should take this into account and repeat the tests conducted ensuring a wider sampling, to review the results and conclusions made.

Above that, the scope of this research has been limited to German consumers. Widening the empirical study in terms of geographical and intercultural outreach to investigate if there exist intercultural differences certainly would lead to relevant insights, especially for international companies considering different regional approaches to CSR communication.

However, future research on the influence of emotions on issues of credibility and trust, can benefit from the developed measurement model for emotions and credibility in this thesis. This model might not only be useful for an extended study on emotions and credibility with different types of CSR content from different firms, countries etc., but it can also be easily adjusted for other areas of communication studies. Wherever the influence of emotions on credibility and trust is of interest, the developed model can be used for further assessment.

7.2. Practical implications

As it was stated earlier in this thesis already, for companies it is no longer enough to have a CSR strategy, but it is essential to effectively communicate this CSR engagement to the public. But how to effectively communicate with the consumer in a constantly changing and emotionalized environment?

For this it crucial, to have a CSR communication that engages with the consumer and other stakeholders, seeking for mutual exchange and consensus. Rather than using short-term marketing or advertising campaigns, it is advisable to have a clear CSR strategy that builds on long-term relationships and exchange with stakeholders. The best way for a company to achieve and maintain credibility is a CSR communication whose main aim is moral legitimacy (see 3.2.). This can be

done by successfully implement the four validity claims, which have been operationalized as dimensions of credibility in this study: understandability, truth, sincerity and normative appropriateness.

The measurement model for emotions and credibility that has been developed in this thesis, can be used by companies to further assess their CSR communication and the effects it has on the consumers’ emotions and the perceived credibility of the firm.

In the analysis, it was shown that emotions and emotional appeal influence the perceived credibility of the consumer in multiple ways. While the high emotional content enhances the understandability, it could be proved that it also diminishes sincerity of CSR communication. Hence, a balance between low and high emotional appeal in CSR content needs to be found.

Both positive and negative emotions influence the perceived credibility of CSR text-image content, with negative emotions having a slightly stronger impact. At the same time, positive and negative emotions showed a high correlation for the text-image content. By aiming towards the creation of positive emotions in CSR text-image content, there is a high risk that unintentionally negative emotions are created and/or intensified. The positive effect on perceived credibility is then nearly diminished. High-emotional CSR text-image content should, thus, be handled with care.

In contrast to that, it is advisable to create more emotional CSR video content, since emotions have a positive overall influence on perceived credibility. By stimulating positive emotions, such as joy, pleasure, contentment etc., the credibility of the firm can be enhanced. This is especially true for interest and compassion, which together had the most influence on credibility. However, the issue of sincerity, which got partly diminished by high emotional content, should still be addressed and taken into consideration when planning CSR video content, i.e. finding a balance between creating positive emotions for the consumer and keeping a genuine communication that avoids appearing biased or being a self-laudation. A successful CSR communication creates interest and positive emotions for the consumer, but is also informative rather than promotional.

Regarding the dimensions of credibility, it has been shown that it is mainly understandability that gets negatively influenced by negative emotions, while positive emotions positively influence truth, sincerity and normative appropriateness. This accounts for a negative influence on one dimension versus a positive influence on three dimensions.

Rather than overcautiously avoiding the creation of emotions because of concerns to stimulate negative reactions, the aim of a successful CSR communication strategy should it, thus, be to create emotions that make the consumer feel positively and interested in the topic of CSR. At least for the CSR communication content under investigation, the effect of positive emotions overall promises more, than should be feared by negative reactions. The results of this study recommend to include positive emotions and interesting and informative content to CSR communication, without worrying to much about potentially negative reactions towards a proactive CSR communication.

8. References

Baker, Stephanie A., and David Rowe. 2013. “The power of popular publicity: New social media and the affective dynamics of the sport racism scandal.” Journal of Political Power, 6 3: 441–60.

Barbalet, J. M. 1998. Emotion, Social Theory, and Social Structure: A Macrosociological Approach. Cambridge: Cambridge University Press.

Bentele, Günter, and Howard Nothhaft. 2011. “Trust and Credibility as the Basis of Corporate Social Responsibility.” In The handbook of communication and corporate social responsibility, edited by Jennifer Bartlett, Steve K. May, and Øyvind Ihlen, 208–230. Chichester, West Sussex: Wiley-Blackwell.

Bentele, Günter, and René Seidenglanz. 2008. “Trust and Credibility – Prerequisites for Communication Management.” In Public relations research: European and International perspectives and innovations, edited by Ansgar Zerfass, Betteke Ruler, and Krishnamurthy Sriramesh, 49–62. Wiesbaden: VS Verlag für Sozialwissenschaften / GWV Fachverlage GmbH Wiesbaden.

Bhattacharya, C. B., Daniel Korschun, and Sankar Sen. 2009. “Strengthening Stakeholder– Company Relationships Through Mutually Beneficial Corporate Social Responsibility Initiatives.” Journal of Business Ethics, 85 S2: 257–72.

Bryman, Alan. 2016. Social research methods, Fifth edition. Oxford: Oxford University Press.

Cannon, Tom. 1994. Corporate responsibility: A textbook on business ethics, governance, environment ; roles and responsibilities. Harlow: Pearson Education.

Carrington, Michal, and Benjamin Neville. 2015. “CSR and the Consumer.” In Corporate social responsibility, edited by Esben R. G. Pedersen, 166–204. Los Angeles Calif. u.a. SAGE Publ.

Carroll, Archie B. 1991. “The pyramid of corporate social responsibility: Toward the moral management of organizational stakeholders.” Business Horizons, 34 4: 39–48.

Couper, Mick. 2008. Designing effective Web surveys. Cambridge: Cambridge University Press.

Damasio, Antonio R. 1994. Descartes' error. G.P. Putnam.

Deutsche Gesellschaft für Psychologie e.V. 2016. “Berufsethische Richtlinien.” Accessed May 11, 2016. https://www.dgps.de/index.php?id=85

Donsbach, Wolfgang, ed. 2008. The international encyclopedia of communication. Oxford: Wiley- Blackwell.

Eilbirt, Henry, and Robert Parket. 1973. “The Current Status Of Corporate Social Responsibility.” Business Horizons, 16 4: 5.

European Commission. 2001. “Green paper - Promoting a European framework for corporate social responsibility.” http://eur-lex.europa.eu/legal- content/EN/TXT/PDF/?uri=CELEX:52001DC0366&from=EN

European Commission. 2011. “A renewed EU strategy 2011-14 for Corporate Social Responsibility.” http://eur-lex.europa.eu/legal- content/EN/TXT/PDF/?uri=CELEX:52011DC0681&from=DE

Farache, Francisca, and Keith J. Perks. 2010. “CSR advertisements: A legitimacy tool?” Corporate Communications: An International Journal, 15 3: 235–48.

Freeman, R. E. 2010. Stakeholder theory: The state of the art. Cambridge u.a. Cambridge Univ. Press.

Frey, Bruno S., Jana Gallus, and Lasse Steiner. 2014. “Open issues in happiness research.” International Review of Economics, 61 2: 115–25.

Frey, Bruno S., and Alois Stutzer. 2005. “Happiness Research: State and Prospects.” Review of Social Economy, 63 2: 207–28.

Frey, Bruno S., Alois Stutzer, and Matthias Benz. 2008. Happiness: A revolution in economics.

FUTERRA Sustainability Communications Ltd. 2005. “The Rules of the Game: Principles of Climate Change Communications.” Accessed 07.04.17. http://www.stuffit.org/carbon/pdf- research/behaviourchange/ccc-rulesofthegame.pdf

Gardner, Meryl P. G., ed. 2015. Emotions. Los Angeles [u.a.]: SAGE Publications.

Giddens, Anthony. 2007. “The globalizing of modernity.” In The global transformations reader : an introduction to the globalization debate, 60–66. Cambridge: Polity [u.a.].

Grappi, Silvia, Simona Romani, and Richard P. Bagozzi. 2013. “Consumer response to corporate irresponsible behavior: Moral emotions and virtues.” Journal of Business Research, 66 10: 1814– 21.

Gravetter, Frederick J., and Larry B. Wallnau. 2009. Statistics for the behavioral sciences, 8. ed., internat. student ed. Belmont, Calif. Wadsworth Cengage Learning.

Haidt, Jonathan. 2012. The righteous mind. Pantheon Books.

Hair, Joseph F., William C. Black, Barry J. Babin, and Rolph E. Anderson. 2014. Multivariate data analysis. Pearson Education Limited.

Hansen, Anders, and David Machin. 2013. Media and communication research methods. Basingstoke: Palgrave Macmillan.

Hermida, Alfred. 2016. “Emotion sells: Donald Trump supporters took over media by placing feelings over facts.” Salon.com, November 20. Accessed February 01. http://www.salon.com/2016/11/20/trump-supporters-became-the-media-by-privileging-emotions- over-facts_partner/

Holbrook, Morris B., and Rajeev Batra. 2015. “Assesing the Role of Emotions as Mediators of Consumer Responses to Advertising.” In Emotions, edited by Meryl P. G. Gardner, 404–420. Los Angeles [u.a.]: SAGE Publications.

ISO. 2010. “Guidance on social responsibility.” https://www.iso.org/obp/ui/#iso:std:iso:26000:ed- 1:v1:en

Jackob, N. 2008. “Credibility effects.” In The international encyclopedia of communication, edited by Wolfgang Donsbach, 1044–1047. Oxford: Wiley-Blackwell.

Lazzarini, Sergio G., and Pedro Carvalho de Mello. 2001. “Governmental versus self-regulation of derivative markets: Examining the U.S. and Brazilian experience.” Journal of Economics and Business, 53 2: 185–207.

Likert, Rensis. 1932. A Technique for the Measurement of Attitudes. New York, NY.

Lock, Irina. 2016. “Glaubwürdigkeit in der CSR-Kommunikation.” Publizistik, 61 4: 413–29.

Mayer, Roger C., James H. Davis, and F. D. Schoorman. 1995. “An Integrative Model of Organizational Trust.” The Academy of Management Review, 20 3: 709¶mdict=sv-SE.

Newell, Stephen J., and Ronald E. Goldsmith. 2001. “The development of a scale to measure perceived corporate credibility.” Journal of Business Research, 52 3: 235–47.

Norman, Geoffrey R., and David L. Streiner. 2003. PDQ statistics. http://search.ebscohost.com/login.aspx?direct=true&scope=site&db=e000xat&AN=432958

Oxford Dictionary. 2017. “Trustworthiness. Credibility.” Accessed March 15. https://en.oxforddictionaries.com/

Parmar, Bidhan L., R. E. Freeman, Jeffrey S. Harrison, Andrew C. Wicks, Lauren Purnell, and Simone de Colle. 2010. “Stakeholder Theory: The State of the Art.” The Academy of Management Annals, 4 1: 403–45.

Pedersen, Esben R. G., ed. 2015. Corporate social responsibility. Los Angeles Calif. u.a. SAGE Publ.

Pedersen, Esben R. G. 2015. “The Anatomy of CSR.” In Corporate social responsibility, edited by Esben R. G. Pedersen, 3–36. Los Angeles Calif. u.a. SAGE Publ.

Pérez, Andrea, and Ignacio Rodríguez del Bosque. 2015. “An Integrative Framework to Understand How CSR Affects Customer Loyalty through Identification, Emotions and Satisfaction.” Journal of Business Ethics, 129 3: 571–84.

Preston, Peter. 2016. “Trust in the media is the first casualty of a post-factual war.” theguardian, September 25. Accessed February 01. https://www.theguardian.com/media/2016/sep/24/trust-in- media-first-casualty-post-factual-war-corbyn-trump

Rasmussen, Jeffrey L. 1989. “Analysis of Likert-scale data: A reinterpretation of Gregoire and Driver.” Psychological Bulletin, 105 1: 167–70.

Raupp, Juliana, Stefan Jarolimek, and Friederike Schultz, eds. 2011. Handbuch CSR: Kommunikationswissenschaftliche Grundlagen, disziplinäre Zugänge und methodische Herausforderungen ; mit Glossar. Wiesbaden: VS-Verl.

Rohrmann, Bernd. 1978. “Empirische Studien zur Entwicklung von Antwortskalen für die sozialwissenschaftliche Forschung.” Zeitschrift für Sozialpsychologie, 9: 222–45.

Röttger, Ulrike, Joachim Preusse, and Jana Schmitt. 2014. Grundlagen der Public Relations: Eine kommunikationswissenschaftliche Einführung, 2., akt. Aufl. 2014. Wiesbaden: VS Verlag für Sozialwissenschaften.

Rouner, D. 2008. “Credibility of content.” In The international encyclopedia of communication, edited by Wolfgang Donsbach. Oxford: Wiley-Blackwell.

Scherer, Klaus R. 2005. “What are emotions? And how can they be measured?” Social Science Information, 44 4: 695–729.

Schwegler, Petra. 2013. “14-59: RTL erweitert werberelevante Zielgruppe nach oben.” W&V, February 28. Accessed April 27. https://www.wuv.de/medien/14_59_rtl_erweitert_werberelevante_zielgruppe_nach_oben

Sen, Sankar, and C. B. Bhattacharya. 2001. “Does Doing Good Always Lead to Doing Better? Consumer Reactions to Corporate Social Responsibility.” Journal of Marketing Research, 38 2: 225–43.

Våland, Terje, and Morten Heide. 2005. “Corporate Social Responsiveness.” European Management Journal, 23 5: 495–506.

Venkatesh, Alladi, and Laurie A. Meamber. 2008. “The aesthetics of consumption and the consumer as an aesthetic subject.” Consumption Markets & Culture, 11 1: 45–70.

Xie, Chunyan, Richard P. Bagozzi, and Kjell Grønhaug. 2015. “The role of moral emotions and individual differences in consumer responses to corporate green and non-green actions.” Journal of the Academy of Marketing Science, 43 3: 333–56.

Appendix

Appendix 1: Design feature differentiation of different types of affective phenomena

Source: Scherer (2005)

Appendix 2: Presentation of the selected CSR communication content

GUT GEMACHT – Das WE Programm

GUT GEMACHT – Nachhaltige Textilien

Nachhaltigkeitsbericht 2014 – Nachhaltige Lieferketten

Nachhaltigkeitsbericht 2014 – Nachhaltige Ressourcen und Produkte

Mary oder Hannah- Wer erzählt von ihrem Schultag

Nachhaltigkeit - Das WE Programm in Vietnam

Nachhaltigkeit – Gemeinsam mit Cotton made in Africa für die Zukunft in Afrika

Nachhaltigkeit - Wir, das heißt gemeinsam handeln!

Appendix 3: The Geneva Emotion Wheel (German)

Empfohlene Instruktion für das Geneva Emotion Wheel

(GEW, Version 3.0)

(BITTE AN DIE JEWEILIGE VERWENDUNG ANPASSEN)

Das folgende Instrument dient dazu, die Emotionen, die Sie empfunden haben, möglichst genau zu erfassen…

(BITTE DETAILLIERTE INSTRUKTIONEN FUER IHRE STUDIE EINFUEGEN)

Die 20 kreisförmig angeordneten Emotionsbegriffe des untenstehenden Emotions-Rades sollen es Ihnen erleichtern, die von Ihnen empfundene Emotion zu beschreiben. Bitte berücksichtigen Sie, dass die Begriffe jeweils eine breitere „Emotionsfamilie“ repräsentieren und damit jeweils auch eine Reihe von ähnlichen Emotionen umfassen. Das Feld „Ärger“ beinhaltet demnach beispielsweise auch Emotionen wie Wut, Gereiztheit, Verdruss, Genervtsein, Empörung usw., das Feld „Angst“ beinhaltet Besorgnis, Beklemmung, Furcht und Panik. Manche Begriffe wie Liebe, Hass oder Schuld können sich auch auf langfristige Gefühlszustände beziehen. Im vorliegenden Fall meinen diese Begriffe jedoch ein momentanes, kurzfristiges Gefühl von Liebe, Hass, oder Schuld in einer bestimmten Situation.

Bitte wählen Sie zunächst den Emotionsbegriff bzw. die Emotionsfamilie aus, der/die am ehesten das Gefühl beschreibt, welches Sie während der Aufgabe empfunden haben. Bitte wählen Sie einen der vorgegeben Begriffe, auch wenn dieser Ihr Erleben nicht vollständig beschreiben sollte. Markieren Sie anschliessend durch Ankreuzen eines der Kreise in der gewählten "Speiche", mit welcher Intensität Sie die entsprechende Emotion empfunden haben – je grösser und je weiter aussen der Kreis, desto intensiver und stärker die Emotion. Verschiedene Intensitäten entsprechen oftmals verschiedenen Emotionsbegriffen innerhalb einer Emotionsfamilie. Beispielsweise kann Gereiztheit als eine weniger intensive Form von Wut angesehen werden; Besorgnis stellt hingegen eine weniger intensive Form von Angst dar. Um solche weniger intensive Emotionen auszuwählen, kreuzen Sie bitte einen der kleineren, näher an der Mitte liegenden Kreise an. Sehr starke Emotionen können Sie durch Ankreuzen der grössten, aussen liegenden Kreise kennzeichnen.

(BITTE WÄHLEN SIE JE NACH FRAGESTELLUNG IHRER STUDIE EINE DER FOLGENDEN INSTRUKTIONEN AUS)

(MÖGLICHKEIT 1 – PROBAND WÄHLT EINE ANTWORT AUS)

Bitte wählen Sie diejenige Emotion in der entsprechen Intensität aus, welche am besten das Gefühl beschreibt, welches Sie soeben empfunden haben. Sollten Sie keine Emotion empfunden haben, kreuzen Sie bitte das Feld „Keine“ in der Mitte des Kreises an. Sollten Sie eine Emotion empfunden haben, die keinem der vorgegebenen 20 Begriffe ähnelt, kreuzen Sie bitte das Feld „Andere“ in der Mitte des Kreises an.

(MÖGLICHKEIT 2 – PROBAND KANN MEHRERE GLEICHZEITIG AUFTRETENDE ODER GEMISCHTE EMOTIONEN WÄHLEN)

Emotionale Reaktionen sind zuweilen sehr komplex und können gleichzeitig Facetten verschiedener Emotionen enthalten. Bitte kreuzen Sie all diejenigen Emotionen in der jeweiligen Intensität an, welche zu dem Gefühl beigetragen haben, das Sie soeben empfanden. Die Intensitäten dieser Emotionen können unterschiedlich stark sein. Sollten Sie keine Emotion empfunden haben, kreuzen Sie bitte das Feld „Keine“ in der Mitte des Kreises an. Sollten Sie eine Emotion empfunden haben, die keinem der vorgegebenen 20 Begriffe ähnelt, kreuzen Sie bitte das Feld „Andere“ in der Mitte des Kreises an.

(MÖGLICHKEIT 3 – PROBAND GIBT EINE EXPLIZITE BEWERTUNG ALLER EMOTIONEN AB)

Emotionale Reaktionen sind zuweilen sehr komplex und können gleichzeitig Facetten verschiedener Emotionen enthalten. Bitte wählen Sie für jede angegebene Emotion aus, mit welcher Intensität diese zu Ihrem Empfinden in der soeben erlebten Situation beigetragen hat. Diese Intensitäten können durchaus sehr gering sein. Hat eine Emotion überhaupt nicht zu Ihrem Gefühl beigetragen, kreuzen Sie bitte das kleine Kästchen unter dem Kreis für die geringste Intensität an.

(WENN SIE ZUSÄTZLICHE INFORMATIONEN ZU EINER NICHT AUFGEFÜHRTEN EMOTION ERHEBEN WOLLEN, BITTEN SIE DEN PROBANDEN, DEN JEWEILIGEN BEGRIFF IN DIE LEERE ZEILE IM FELD „ANDERE“ ZU SCHREIBEN).

Wut Interesse

Hass Belustigung

Verachtung Stolz

Ekel Freude

Vergnügen Angst Keine

Andere Zufriedenheit Enttäuschung ______

Liebe Scham

Bereuen Bewunderung

Schuld Erleichterung

Trauer Mitgefühl

Appendix 4: The Geneva Emotion Wheel (English)

Geneva Emotion Wheel Version 3.0

Instructions

(TO BE ADAPTED ACCORDING TO THE INTENDED USE)

This instrument, called the Geneva Emotion Wheel, is used to measure as precisely as possible the emotion you experienced ... (INSERT DETAILED INSTRUCTIONS FOR YOUR STUDY HERE)

In order to make it easier for you to report the type of emotion you experienced, 20 different emotions are arranged in a circular fashion on the following response sheet. Please note that the words provided often represent a large "emotion family" and may thus refer to a whole range of similar emotions. Thus, the Anger family also covers emotions such as rage, vexation, annoyance, indignation, fury, exasperation, or being cross or mad; the Fear family includes anxiety, worry, apprehensiveness, fright, or panic. Some of the words, such as love, hate, or guilt, can be used to refer to long-term affective states; but in this case checking those labels means that you have had a salient temporary feeling that belongs to the families of Love, Hate, or Guilt.

First identify approximately what the event that produced the emotion meant to you and choose the emotion family that seems to best correspond to the kind of feeling you experienced when this happened, even though the words on the sheet may not capture all facets of your experience. Then determine with which intensity you experienced the respective emotion and check one of the circles in the "spike" corresponding to this emotion family -- the bigger the circle and the closer it is to the rim of the wheel, the stronger your emotional experience would have been. Different intensities often correspond to different members of an emotion family. Thus, irritation can be considered a less intense emotion belonging to the Anger family and anxiety a less intense emotion belonging to the Fear family. For less intense emotions, please check one of the smaller circles in the spike. If the emotion was very intense, please check the largest circle of the spike.

(CHOOSE ONE OF THE FOLLOWING RESPONSE FORMATS ACCORDING TO YOUR RESEARCH AIMS)

(ALTERNATIVE 1 – PARTICIPANT CHOOSES ONE EMOTION ONLY)

Please rate the intensity of the emotion in the wheel which best describes what you felt. If you did not feel any emotion at all, please check the upper half circle in the center of the wheel (labeled "None"). If you experienced an emotion that is very different from any of the emotions in the wheel, please check the lower half circle (labeled "Other").

(ALTERNATIVE 2 – ALLOW CHOICE OF SEVERAL SIMULTANEOUS EMOTIONS OR BLENDS)

Our emotions are sometimes blended or mixed and may contain many different components. Please rate the intensity of those emotions in the wheel that contributed to what you felt, even if the intensities are different. If you did not feel any emotion at all, please check the upper half circle in the center of the wheel (labeled "None"). If you experienced an emotion that is very different from any of the emotions in the wheel, please check the lower half circle (labeled "Other").

(ALTERNATIVE 3 – FORCED RATING OF ALL EMOTIONS IN THE WHEEL, IN CASE EXPLICIT RATINGS ARE REQUIRED FOR ALL EMOTIONS)

Our emotions are often blended or mixed and contain many different components. Please rate the intensity of all the emotions in the wheel with respect to what you felt, even if the intensities are very low. For those emotions that were not at all part of your reaction, please check the small box under the smallest circle, respectively.

(IF YOU WANT RESPONDENTS TO REPORT EMOTIONS THEY FELT AND THAT ARE NOT LISTED ON THE WHEEL YOU CAN TELL THEM IN YOUR INSTRUCTIONS TO WRITE THE RESPECTIVE WORD OR EXPRESSION IN THE SPACE PROVIDED IN THE "OTHER" HALF CIRCLE IN THE CENTER OF THE WHEEL.

Anger Interest

Hate Amusement

Contempt Pride

Disgust Joy

Pleasure Fear None